There’s no shortage of people breaking down how companies make money. Revenue curves, funding rounds, ARR milestones. I read a lot of them.

What almost nobody breaks down is the layer underneath:

• How these companies won attention in the first place.

• Which channel they started on.

• Which format actually broke through.

• How often they published, and

• What they eventually killed.

That’s the part that decides whether any of the revenue happens, and it’s the part that gets skipped.

So I’m building it.

What this is: I take one company apart through a content lens and answer all the questions you may ask about that company’s growth…especially how they used content to drive their growth, and I will provide real numbers…

What you get: how they built their audience, plus what the company stopped doing, (that means what’s working for them) and what won’t transfer to you if you’re one person without a team or a budget.

Where this is going: With this series, I am creating a filterable database. Filter by channel, by format, by team size, by whether a solo creator can realistically pull it off…each entry linking back to the full issue. So instead of trying to remember which breakdown had the play you need, you just filter for it.

Let’s start with a company that grew a blog to 1.5M monthly visits by publishing less.

Subscribe now

📊 The Scoreboard

Primary channel Organic search (blog), then YouTube

Time to first traction ~12 months of failure, then 4 years of compounding

Content team at breakout Small — a handful of marketers

Publishing cadence Cut from 3 posts/week to ~1, later ~2

Peak measurable outcome 15K → 1.5M+ monthly blog visits · 10K → 600K YouTube

Company today ~141 employees, $100M+ ARR, bootstrapped, no sales team

Ahrefs doesn’t run Google Analytics on its blog.

It doesn’t put a lead magnet on the page. Doesn’t run an exit popup. Doesn’t track how many blog readers convert into customers. The company that sells the software marketers use to measure content has, by choice, almost no measurement on its own content.

And it worked. The blog went from about 15,000 monthly visits to over 1.5 million. The company crossed $100M ARR without taking a dollar of venture money and, reportedly, without a sales team.

Here’s the part that should bother you more, though.

When that blog was stuck at 15,000 visits, they were publishing three articles every week. The traffic didn’t move. So they cut back to roughly one post a week, and traffic grew tenfold.

Everything you’ve been told about content volume is in that sentence, backwards.

What you’ll learn:

• The 0–3 score that decides every topic Ahrefs writes about

• Why publishing less was the unlock, not the compromise

• The argument for collecting zero emails from a million-visitor blog

• How one landing page went from position 8 to position 1 by becoming something else

• What Ahrefs killed in 2025 when the ground moved under them

Method: Built from 12+ founder and operator interviews, podcasts, and Ahrefs’ own published posts, plus public company data. Figures Ahrefs hasn’t disclosed are third-party estimates and labeled as such. Sources at the bottom.

Subscribe now

The Setup

Ahrefs was founded in July 2010 by Dmitry Gerasimenko, a Ukrainian developer. Bootstrapped. No investors, ever. The company later moved its HQ to Singapore.

By 2015 they had a real product and a blog that wasn’t working. Around 15,000 visits a month, flat, despite a three-posts-per-week schedule. That’s the year Tim Soulo joined to run marketing.

His first move was the obvious one: publish better content, more consistently. He did that for a year.

It produced almost nothing.

The content was genuinely good, that wasn’t the problem. It just wasn’t aligned to anything people were actually searching for. Which is a particularly uncomfortable thing to discover at a company whose entire product exists to tell you what people are searching for.

That failed year is the most useful part of this whole story, and it’s the reason I’m starting the series here.

The fix wasn’t effort. They were already trying hard. The fix was a filter.

The Seven Content Plays

Play 1 — The Wedge Channel

Ahrefs picked organic search and stayed there. Not as a phase, as the entire strategy, for years, before YouTube got serious attention.

The logic is almost embarrassingly aligned: they sell an SEO tool, so their buyers are people who search. Every article that ranked put the product in front of someone already trying to solve a problem the product solves.

Soulo has been blunt that SEO isn’t universal, if there isn’t enough search volume in your niche, it’s the wrong channel and you should pick a different one. The wedge wasn’t “SEO is best.” It was “SEO is where our buyers already are.”

Most creators pick a wedge based on where they’re comfortable. Ahrefs picked based on where the buyer was standing.

Play 2 — The Breakout Format

The format is a tutorial that demonstrates the tool while solving the reader’s problem.

Not a thought-leadership essay. Not a trend piece.

A how-to, targeting a search query, where the screenshots showing the solution happen to be screenshots of Ahrefs.

Underneath it sits the framework worth stealing…the Business Potential score. Every topic gets rated 0 to 3:

• 3 — the product is an irreplaceable solution to this problem

• 2 — the product helps a lot, but isn’t essential

• 1 — the product can be mentioned, barely

• 0 — no natural connection

They only write 2s and 3s. Almost zero flexibility, no chasing trends, no covering topics because someone found them interesting.

This is why the blog converts without a funnel. There is no gap between “reader has a problem” and “product solves it” for the funnel to sit in.

Play 3 — The Cadence Reality

Three posts a week produced a plateau. In 2016–17 they dropped to roughly one post a week. Traffic went up tenfold over the following years, later settling around two a week.

Read that again, because it inverts the advice every creator gets.

The volume wasn’t the constraint. The filter was. Three mediocre-fit posts a week generate less than one high-fit post a week, because search compounds on quality of match, not quantity of output.

There’s a second layer here.

Soulo has said his primary KPI was never traffic….it was ARR. Traffic is a number you can grow by writing about anything. He deliberately refused to optimize for the metric that was easiest to move.

Team size matters for honesty: this was a small marketing team, not a content factory.

The output was low because the standard was high, and they could afford that standard because they weren’t chasing a traffic target.

Play 4 — The Distribution Loop

Someone searches a problem → finds an Ahrefs tutorial → the tutorial solves it using Ahrefs → they sign up → they need to learn the tool → they search “how to do X with Ahrefs” → they find another Ahrefs tutorial.

The same article does acquisition and retention. It ranks in Google for a stranger and educates a paying customer, and the education reduces churn while generating word of mouth. One asset, three jobs.

Distribution after publishing was unglamorous and multi-channel: email list, Twitter, LinkedIn, Facebook groups, some paid retargeting, and an in-app notification bell inside the product itself.

That last one is the quiet advantage…they could put every new article in front of their entire customer base without paying anyone.

Play 5 — The Capture Point

This is the play that breaks the template, and it’s the most interesting thing Ahrefs does.

They don’t capture emails on the blog. No lead magnets, no gated content upgrades, no popups.

Soulo’s reasoning: a reader on an Ahrefs article is already on a perfectly tailored sales page. Interrupting them with an email capture doesn’t move them closer to buying, it moves them sideways into a nurture sequence they didn’t ask for, and it makes the page worse for the person reading it.

So the capture point isn’t an email form. It’s the free product tier. Ahrefs Webmaster Tools gives verified site owners a full Site Audit and limited Site Explorer data for their own domain, at zero cost.

You don’t join a list. You start using the thing.

Whether this is correct is genuinely arguable…see the transfer section below. But it’s a real position, held deliberately, for a decade.

Play 6 — The Repurposing System

The YouTube channel started about as unglamorously as possible. Sam Oh joined in a content role and began taking existing blog posts, writing scripts from them, and turning them into videos.

No new research. No separate content strategy. The same Business Potential filter, applied to a second surface.

That channel went from roughly 10,000 subscribers to over 600,000. The discipline carried across: Oh has described a ratio of roughly nine-to-one in favor of videos where the tools can actually be demonstrated. Videos were built into deep playlists rather than one-offs…dozens of videos on a single theme, so a beginner could binge an entire education in one sitting.

Play 7 — The Compounding Asset

The clearest single example in this entire breakdown.

Ahrefs had a feature landing page for its backlink checker. It was stuck around position 8 in Google, because everyone ranking above it offered a free tool while Ahrefs offered a page with a trial button on it.

So they turned the landing page into the free tool.

It went to position 1.

That’s the whole lesson, and it generalizes past SEO: when the search intent is “let me try something,” a page describing the thing will always lose to the thing.

The free tool suite that grew out of this now functions as the top of the entire funnel, and it keeps working with no ongoing content cost.

Every durable content company I’ve looked at has exactly one of these. For Ahrefs it’s free tools.

The Kill List

👉 They killed the three-posts-a-week schedule. The single highest-leverage decision in the story, and it was a decision to do less.

👉 They killed topic freedom. No trend-chasing, no writing about whatever’s interesting this week. Soulo has described it as having almost no flexibility in what they write about…a constraint they imposed on themselves.

👉 They killed traffic as a KPI. ARR was the number. Traffic that didn’t convert was, correctly, treated as vanity.

👉 They killed the “great content is enough” belief. Soulo tells a story about spending a month on a piece he thought was his best work, sending it to Rand Fishkin for a share, and getting told it didn’t say anything new. Quality without novelty and without search demand is just an expensive hobby.

👉 And in 2025–26, they started killing pure search dependence. AI Overviews cut click-through rates hard, Ahrefs’ own research reported roughly a 35% drop across 300,000 keywords, and clicks to the number one result falling further from there. Their response was to diversify into LinkedIn, YouTube, original data studies, a conference, and brand-led work, including comparison campaigns willing to make fun of themselves. Their CMO’s framing now is that citations are the new backlinks, and being mentioned matters more than being ranked.

The company that won by going all-in on one channel is currently, publicly, in the middle of leaving it.

What Doesn’t Transfer

The no-email-capture play. This works because Ahrefs’ product is the literal answer to the search query. A reader is one click from the solution. If you’re a creator selling a $97 course, your reader is not one click from the solution…they need a relationship first, and refusing to build an email list would be self-harm. Copy the reasoning, not the tactic: don’t interrupt someone who’s already at the point of purchase. That’s different from never capturing anyone.

The in-app notification bell. You don’t have one. That’s a distribution channel available only to companies with a product base, and it quietly did a lot of work here.

The patience. A year of failure and four more years of compounding is a five-year horizon, funded by an existing profitable product. Most solo creators cannot fund five years of anything.

The Business Potential score, however, transfers completely. It requires no budget, no team, and no traffic. It’s just a filter, and it’s the piece I’d argue is doing most of the work.

Steal from This

1. Score your next ten content ideas 0–3. (from Play 2) How irreplaceable is your offer as the solution to that topic? Delete everything scoring 0 or 1. If you have no offer yet, that exercise will tell you something uncomfortable and useful. Also, let me share something I have developed to create a solid offer

2. Cut your publishing cadence in half and put the time into the filter. (from Play 3) If Ahrefs went from 3 posts a week to 1 and grew tenfold, your posting frequency is almost certainly not your bottleneck.

3. Find your one page that should be a tool. (from Play 7) Something you currently describe that you could let people use…a calculator, a template, a checklist, a mini-audit. Descriptions lose to things.

Which of these seven plays is the one you’re avoiding? Hit reply…I read every one.

Keep it simple,

Sharyph
Founder of The Digital Creator

Sources

Primary — Ahrefs and its team

1. Tim Soulo — How We Grew Traffic to Ahrefs’ Blog by 1136% (Medium) — cadence, Business Potential, ARR as KPI

2. Ahrefs Blog — The Great Diversification — AI Overviews click data, channel diversification

3. Ahrefs — Webmaster Tools — the free tier

4. Sam Oh — personal site — YouTube growth from ~10K to 600K subscribers

5. Tim Soulo — personal blog — Business Potential, $100M ARR reference

Interviews and podcasts

6. SaaS Club — SaaS Growth & Unconventional Wisdom with Tim Soulo — no analytics, no lead capture, backlink checker story

7. The Marketing Meetup — Ahrefs’ blog to 1.5M visits per month — traffic milestones, SEO isn’t universal

8. Directive — Yours in Marketing Ep. 5 with Tim Soulo — Business Potential explained

9. Databox — Growing Customers & Revenue with Tim Soulo — the 0–3 scale in detail

10. Upraw Media — What many companies get wrong about SEO — the Rand Fishkin story

11. Matt Byrom — Marketing Strategies Podcast with Tim Soulo — small team, word of mouth

12. Growth Marketing Today — How Ahrefs Grew Their YouTube Channel with Sam Oh — video strategy

13. SEO for the Rest of Us — YouTube SEO Strategies with Sam Oh — the 9:1 product-demonstration ratio

14. Siege Media — Inside Ahrefs’ $100M Growth (also on YouTube) — 2025 diversification

15. SearchPilot — From SEO to GEO in 2026 with Tim Soulo — brand campaigns, AI-era strategy

16. The FXCK — Tim Soulo on organic lead strategy — product-led tutorial format

Third-party analysis (estimates, not company-confirmed)

17. Product Growth — Ahrefs company profile — founding date, headcount, ARR, bootstrapped status

18. The B2B Creator — Ahrefs’ video marketing playbook — how the YouTube channel started

19. SwayOne — How Ahrefs Grew to 600,000 YouTube Subscribers — playlist and series structure

20. BloggersIdeas — Tim Soulo on AI search and brand mentions — citations as the new backlinks

Welcome to The Digital Creator, helping creators build better businesses with better systems.

TL;DR: Stop asking "what should I sell?" Your existing content…newsletters, posts, videos, repeated questions…already contains the problems, expertise, and proof you need. Extract the 10 hidden signals, find the recurring pattern, turn it into a specific transformation, then build the mechanism, package, and proof around it.

Before we get into today’s playbook, I want you to do one thing.

Stop trying to come up with your next offer.

Because here’s what I’ve learned after two years of building in public: you probably already have the raw material for your next product. It’s sitting inside your newsletters. Your posts. Your videos. Your threads. Your frameworks. Your client conversations. Your repeated opinions. Your answers to the same questions you’ve been asked ten times.

You just haven’t organized it yet.

The creator economy grew from $191.55B in 2024 to a projected $234.65B by the end of 2026. And the whole game has shifted. It’s no longer about being a “creator.” It’s about becoming a creator-entrepreneur…someone who owns products, not just an audience.

But most creators are stuck at the starting line. Not because they lack ideas. Because they’re looking in the wrong place for them.

Let me show you where to actually look.

Subscribe now

What If Your Next Offer Is Already Hidden in Your Content?

Most creators start offer creation backwards…picking a product format first, then hunting for a problem to fit it. Your published content already contains accumulated market intelligence: proven problems, developed opinions, and topics your audience responds to. The opportunity isn't inventing something new…it's mining what you've already created.

Most creators approach offer creation backwards.

They start with the product.

“Should I make a course?”

“Should I build a SaaS?”

“Should I create a community?”

“Should I sell consulting?”

Then they try to find a problem that fits the product.

That’s backwards.

Read that again.

Your existing content is a much better starting point. Because you’ve already spent months…maybe years…publishing ideas.

You’ve already discovered what you know.

You’ve already explained problems.

You’ve already developed opinions.

You’ve already answered questions.

And if you’ve been paying attention, you’ve already seen which topics people actually respond to.

Here’s the truth:

Your content is not just marketing.

It’s accumulated market intelligence.

The opportunity isn’t to create something new.

It’s to learn how to mine what’s already there.

Why Most Creators Struggle to Create Offers

Three mistakes kill offer creation before it starts: starting with the product instead of the problem, starting with what you know instead of what transforms someone, and copying a competitor's offer instead of using your own evidence. The fix for all three is the same…start with your content, your audience, and your proof.

I’ve made every one of these mistakes myself.

When I started, I had ideas. But no direction. No clear offer. No system. Just noise.

Here are the three mistakes I see on repeat.

❌ Mistake #1: Starting with the product

You decide to create a $99 course.

Then you spend six weeks trying to figure out what the course should teach.

See what happened? The product became the constraint.

Instead: Start with the problem. Then decide what product is the best vehicle for solving it.

❌ Mistake #2: Starting with what you know

This one sounds smarter.

“What am I good at?”

But even this isn’t enough.

The market doesn’t pay you simply because you know something. It pays you when your knowledge produces a valuable transformation.

Knowing how to write isn’t an offer.

Helping a founder turn their expertise into a content system that generates qualified conversations? That might be.

The distinction matters more than you think.

❌ Mistake #3: Copying someone else’s offer

You see someone selling a newsletter course.

It works for them.

So you make a newsletter course.

But you don’t have their audience.

You don’t have their proof.

You don’t have their positioning.

And you don’t necessarily solve the same problem.

👉 The better strategy is to start with your evidence.

Your content. Your expertise. Your audience. Your problems. Your proof.

That’s where the opportunity lives. Not in someone else’s Gumroad page.

What Should You Actually Look For in Your Content?

Don’t ask “could I sell this piece?”…extract the intelligence underneath it. Look for 10 signals across your content: audience, problems, desires, beliefs, questions, expertise, frameworks, proof, stories, and outcomes. One post is interesting; twenty posts repeating the same problem is a pattern worth building a business around.

Here’s where most people get it wrong.

They read each piece of content and ask:

“Could I sell this?”

That’s too narrow.

Instead, extract the intelligence underneath the content.

For every piece, look for these 10 signals:

1. Audience → Who is this actually for?

2. Problems → What problem are they experiencing?

3. Desires → What do they want instead?

4. Beliefs → What do they currently believe about the problem?

5. Questions → What are they trying to figure out?

6. Expertise → What do you repeatedly teach better than most people?

7. Frameworks → What systems or processes have you developed?

8. Proof → What results, examples or case studies support your ideas?

9. Stories → What experiences make your knowledge credible?

10. Outcomes → What changes for someone who applies what you teach?

The goal isn’t to find one interesting idea.

It’s to find patterns.

One post about content distribution is interesting.

Twenty pieces about distribution, creator burnout, repurposing and content systems? That’s a signal.

That’s where you start paying attention.

💡 Once you've spotted a pattern worth pursuing, a tool that extracts your offer from content you've already published can turn it into positioning, pricing, and email scripts in minutes instead of doing this pass by hand.

How Do You Turn a Content Pattern Into an Offer?

Follow the chain: content reveals a pattern, the pattern points to a real problem, the problem implies a desired outcome, and the outcome becomes your opportunity. A topic like “content repurposing” isn’t an offer…an offer names a specific person, problem, outcome, and mechanism, not just a subject.

From pattern to problem to opportunity

Here’s the framework I use.

Content → Pattern → Problem → Outcome → Opportunity

Let me walk you through a real example.

Say you’ve published several pieces about content creation.

You notice you keep coming back to one thing: creators struggling to distribute their work.

That’s the pattern.

Now investigate further.

The actual problem isn’t:

“Creators need more content.”

It’s:

“Creators are spending hours creating content but aren’t getting enough mileage from the ideas they’ve already produced.”

Now you have a problem.

The desired outcome becomes:

“Get more reach from the content I’m already creating — without spending hours manually rewriting everything.”

Now you have a transformation.

And that creates an opportunity:

A system that turns one strong content pillar into platform-native distribution assets.

Notice what just happened.

We didn’t invent the business idea.

We extracted it.

That’s the whole shift.

Your Topic Is Not Your Offer

This is one of the most important distinctions in this entire playbook.

“Content repurposing” is a topic.

“AI content creation” is a topic.

“LinkedIn growth” is a topic.

“Email marketing” is a topic.

None of those are offers.

An offer is much more specific. It connects:

A person → with → a painful problem → to → a valuable outcome → through → a believable mechanism.

Here’s the formula I use. Steal it:

For [specific person] who [specific problem], [offer] helps them achieve [specific outcome] through [unique mechanism], without [major sacrifice].

Now compare these two:

❌ “A content repurposing course.”

✅ “A system that helps solo creators turn one strong piece of content into a week’s worth of platform-native distribution — without spending hours manually rewriting it.”

The second one tells you who it’s for. What it solves. What changes. And why it’s different.

The difference isn’t subtle.

One is a topic. One is an offer.

The Mechanism Is Where Your Offer Gets Interesting

Two businesses can promise the same outcome.

But the mechanism can be completely different.

And that matters…because buyers don’t just need to want the outcome.

They need to believe your way of getting there actually works.

Here’s what that looks like:

Problem → Creators don’t have enough time to distribute their ideas.

Old mechanism → Create one piece → manually rewrite it five times → publish everywhere.

Why it fails → Too much effort. Generic output. Inconsistent execution. Burnout.

New mechanism → One strong pillar → extract core ideas → adapt each idea to its native format → evaluate the output → repair the weak pieces → distribute.

Outcome → More distribution from the same intellectual work.

That’s a mechanism.

And a strong mechanism gives your offer a reason to exist.

Without it, you’re just another course competing on price.

How Do You Package and Price the Offer?

Value goes up when you increase the desired outcome and the buyer’s perceived likelihood of achieving it, and goes down when you decrease time delay and effort. Build the product around the transformation and mechanism, not the module count…templates, support, and community only belong in the offer if they measurably speed up or de-risk that outcome.

How valuable is the transformation?

Here’s another trap.

Creators try to make an offer more valuable by adding more stuff.

More modules. More templates. More bonuses. More PDFs. More prompts. More videos.

But here’s the thing:

More stuff doesn’t automatically mean more value.

A better way to think about value:

↑ Increase:

• Desired Outcome

• Perceived Likelihood of Achievement

↓ Decrease:

• Time Delay

• Effort & Sacrifice

The ideal offer helps someone achieve something they really want, makes them believe they can achieve it, gets them there faster, and removes as much unnecessary work as possible.

That’s why a simple service can be worth more than a 40-hour course.

The customer isn’t buying the hours of content.

They’re buying the transformation.

Price on the transformation. Not the time you put in.

How Do You Package the Offer?

Once you know the transformation and the mechanism, you build the product around them…not the other way around.

Start with the core transformation.

Then ask one question:

What does the customer actually need to achieve it?

That might include:

• Templates

• Software

• Training

• Checklists

• Examples

• Community

• Support

• Implementation

• Feedback

• Done-for-you work

But every single component needs a job.

Don’t add a bonus because you need something to fill the sales page. Add it because it improves the transformation.

Here’s the stack I use:

Core Transformation

↓

Unique Mechanism

↓

Product / Deliverables

↓

Implementation Support

↓

Risk Reversal

↓

Price

That’s the offer.

Not the number of modules.

💡 If you want a broader map of which specific products fit which stage of a creator business, my digital product playbook walks through the idea-to-income path in more depth.

How Do You Build Proof Before You Have Any?

Never let your promise outrun your proof. Climb five levels in order: product proof (it works), customer proof (others used it), outcome proof (it produced results), comparative proof (it beats alternatives), and authority proof (why you). With zero testimonials, offer a discounted “founding buyer” deal for honest feedback…one proof point is enough to start the momentum

This is where I break from a lot of offer frameworks.

They tell you to make a big promise. Then figure out the proof later.

I do the opposite.

Look at your evidence first.

There are five levels of proof:

1. Product proof → Does the thing actually work?

2. Customer proof → Have other people used it?

3. Outcome proof → Did it produce the promised result?

4. Comparative proof → Can you show your approach works better than the obvious alternatives?

5. Authority proof → Why should someone trust you to solve this?

Now here’s the important part.

If you don’t have outcome proof yet…don’t pretend you do.

Reduce the claim. Run a beta. Get users. Document the process. Create product proof. Then build toward stronger claims.

I started with zero testimonials on my first product. Zero.

So I offered a “founding buyer” deal to a handful of people…free or steep discount…in exchange for honest feedback. Even a single line of proof moved the needle. Once the first few came in, momentum picked up.

Your promise should never outrun your proof.

Should You Even Build This?

Before building, stress-test the opportunity: is the problem painful, frequent, and already being paid for, and is your mechanism meaningfully different? Then answer the harder 2026 question…why should someone buy this instead of just asking ChatGPT or Claude? If your edge isn’t proprietary data, workflow, or implementation, it’s not a real offer yet.

The offer stress test

This is the step almost everyone skips.

They find an idea. They get excited. They start building.

Three months later, they discover nobody wants it.

Don’t do that.

Before you build, score the opportunity. Run it through these questions:

✅ Is this actually painful?

✅ Does the problem happen frequently?

✅ Is solving it worth money?

✅ Are people already spending money to solve it?

✅ Is there urgency?

✅ Can the outcome be clearly explained?

✅ Is the mechanism meaningfully different?

✅ Why should I be the person solving this?

✅ Do I have enough proof?

✅ Can someone understand the offer in one sentence?

And then the question that matters more every single month:

Why this instead of ChatGPT or Claude?

If your answer is:

“Because my prompts are better.”

That’s probably not enough anymore.

Your real differentiation is more likely to be:

• Proprietary data

• Workflow

• Accumulated context

• Specialized knowledge

• Integration

• Evaluation

• Distribution

• Implementation

• Accountability

• A better end-to-end system

That’s what people pay for in the AI era. Not information they can get for free.

💡 Want this scored instead of eyeballed? I built a Claude Skill that scores any offer across five dimensions in about 60 seconds.

When the right answer is “don’t build this”

Now the counterargument. Because if you take this too far, you’ll start seeing a business in everything.

That’s not the goal.

Not every topic deserves an offer.

Sometimes the correct answer is:

Don’t build this.

Maybe the problem isn’t painful enough.

Maybe people don’t have buying power.

Maybe the market’s too crowded.

Maybe you don’t have a real advantage.

Maybe you don’t have enough proof.

Maybe the outcome isn’t measurable.

Or maybe…honestly…you just don’t want to spend the next three years solving this problem.

That’s fine.

The goal isn’t to monetize every idea you’ve ever had.

It’s to find the few opportunities worth building around.

Your content gives you possibilities.

The market decides which ones survive.

💡 Don't have a system yet for any of this?

I put the foundation of this exact approach inside my Lean AI-Powered Business Playbook for Creators — the free lead magnet for people who want the systems-first path without the tool overwhelm. → Grab it here.

Then come back and finish this, because the next part is the whole system.

What’s the Full Content-to-Offer System?

The complete system runs nine steps: ingest your existing content, extract the 10 signals, cluster the recurring patterns, generate multiple opportunities, score them, develop the transformation and mechanism, validate with real buyers before building, launch around the same problem, then learn from every customer interaction…which feeds back into step one.

Here’s the operating system I’d use, start to finish.

01 — Ingest

Start with your existing content.

Newsletters. Posts. Videos. Threads. Articles. Transcripts. Client material.

Everything you’ve already made is raw material.

02 — Extract

Pull out the 10 signals:

Audience. Problems. Desires. Expertise. Beliefs. Frameworks. Stories. Proof. Outcomes.

03 — Cluster

Find the recurring patterns.

What problems keep appearing? What topics keep coming back? What do people repeatedly ask about? What do you repeatedly teach?

04 — Generate

Turn the strongest patterns into several possible opportunities.

Don’t fall in love with the first idea. Create options.

05 — Score

Evaluate each opportunity:

Pain. Demand. Urgency. Differentiation. Credibility. Proof. Monetization. Outcome clarity.

06 — Develop

Take the strongest opportunity and build:

The transformation. The mechanism. The product. The pricing. The risk reversal. The proof strategy.

07 — Validate

Do NOT build the whole thing first.

Test the promise. Talk to buyers. Sell a beta. Create a prototype. Look for real behavior — not polite encouragement.

08 — Launch

Then build your marketing around the same problem and mechanism.

Teach it. Demonstrate it. Show the process. Show the proof. Invite people into the offer.

09 — Learn

Every customer interaction gives you more information.

→ That information improves the product
→ which creates better results
→ which creates better proof
→ which creates better content
→ and the cycle continues.

The Interesting Part Is What Happens Next

Once you understand this system, content creation changes.

You’re no longer publishing just to get views.

You’re collecting intelligence.

A newsletter isn’t just a newsletter → it’s a signal.

A thread isn’t just a thread → it’s a test.

A reader’s question isn’t just a question → it’s potential demand.

A framework that lands isn’t just content → it could be a product.

A problem you keep writing about isn’t just a topic → it could be a business.

That’s the shift:

Your content doesn’t just promote your business. It helps you discover what your business should be.

Read that again.

Frequently Asked Questions

Do I need a large audience before doing this?

No. A smaller audience can actually give you better qualitative signals, because you can talk to people directly and get honest, specific feedback instead of vague reactions. The thing that matters isn’t audience size — it’s the quality of the problem you’re solving and how clearly you can validate it.

What if my content covers a lot of different topics?

That’s exactly why clustering matters more than any single post. Don’t judge individual pieces in isolation — look for recurring patterns across your whole body of work. A scattered archive on the surface usually still circles back to two or three problems you keep addressing from different angles.

What if I don’t have testimonials?

Start with product proof instead of waiting for customer proof. Build a small version, run a beta, and document what happens. Offer a discounted “founding buyer” deal in exchange for honest feedback. Then collect customer and outcome proof as you go — that’s exactly how this approach was built and tested.

What if AI can already solve the problem?

Then your offer has to answer a harder question: why should someone buy your system instead of just asking ChatGPT or Claude directly? That’s where mechanism, accumulated context, workflow, proof, implementation, and specialization become your real edge — information alone stopped being enough to charge for.

How long does it take to turn content into an offer?

Faster than building from scratch, because you’re extracting instead of inventing. Auditing your last 20 pieces for the 10 signals and finding the recurring pattern takes a few hours. Turning that pattern into a validated, priced offer — through scoring, developing, and testing with real buyers — typically takes 1-2 weeks.

Key Takeaways

• Audit your last 20 pieces of content for the 10 signals (audience, problem, desire, belief, question, expertise, framework, proof, story, outcome) instead of asking “could I sell this?”

• Find the pattern, not the idea: one post about a topic is noise; the same problem surfacing across 15-20+ pieces is a validated signal worth building around.

• Name the offer with the formula: for [specific person] who [specific problem], [offer] achieves [specific outcome] through [unique mechanism], without [major sacrifice].

• Climb the proof ladder in order…product, customer, outcome, comparative, authority…and start with a discounted “founding buyer” deal if you have zero testimonials.

• Validate before building: run the 9-step system (Ingest → Extract → Cluster → Generate → Score → Develop → Validate → Launch → Learn) and get real buyer behavior, not polite encouragement, before you build anything.

The Bigger Idea

You don’t need another business idea.

You need a better way to look at the business intelligence you’ve already accumulated.

Your content contains:

Problems. Desires. Expertise. Beliefs. Frameworks. Proof. Demand signals.

The job is simple to say and hard to skip:

Turn those signals into an opportunity.

Turn the opportunity into a transformation.

Then build the offer around the transformation.

That’s how you stop guessing what to sell.

And start discovering what your audience is already telling you to build.

The One Thing to Do This Week

Take your last 20 pieces of content.

Ask one question: “What problem do I keep talking about?”

Don’t look for an offer yet. Just find the pattern.

That pattern is the beginning of your next product.

Whenever You’re Ready to Build It

I’ve spent the last two years turning my own content into products…validating, packaging, pricing, and selling them with a lean AI-powered system instead of a bloated tool stack.

Everything in this playbook…the extraction, the mechanism, the value equation, the proof ladder, the validation…is exactly what the tool I have created does.

👉 Get the AI Offer Crafter

And provide the links of your best posts/newsletters, LinkedIn posts..etc and start from there.

Your content already did the hard part. Let’s turn it into income.

Start with AI Offer Crafter 🚀

Until next time,

Sharyph | Founder of The Digital Creator

Welcome to The Digital Creator, helping creators accelerate with AI. I show you how to use AI to start, create, grow, and scale faster.

Since you’re here, tools I’ve built before:

🚀 The SEO Audit Agent

🚀 The Subscriber Pitch Agent

🚀 My AI Repurposing Tool

Subscribe now

How many hours have you spent scrolling back through your own posts, trying to remember what actually got people to reply?

If you’ve done it, you know the drill:

• Scrolling months back through your own content, hunting for the post that hit

• Trying to recall which question people asked you more than once

• Guessing at a price with literally no anchor

• Writing a DM that feels like begging

• Staring at a blank page, trying to name the thing you’d even sell

You can lose a whole weekend doing this. I know, because I’ve done it more times than I’d like to admit.

And here’s the part that actually stings: even after all that digging, you still can’t answer the only question that matters…would this actually sell? You’ve got a folder of half-ideas and zero conviction.

You’re guessing with better lighting.

Here’s the thing: I’ve actually caused this exact problem before.

Last week I wrote The Offer Playbook and told you to go find your repeated question…the thing you keep explaining for free, scattered across everything you’ve published. A bunch of you did it. You replied with your version of it, exactly like I asked.

And then I watched how long the next step actually took people. Not finding the question…turning it into an offer. A name, a price, a page, an outreach message that doesn’t feel like begging. That part ate weekends.

I’d actually tried to fix this once before. Back in December I built an AI offer analysis tool that scored any offer out of 50...audience specificity, transformation clarity, unique mechanism, the works. It was useful. It told me my positioning was an 18/50 or a 42/50.

But it never once told me what the offer should be. It could grade what I already had. It couldn’t find what I hadn’t named yet.

That’s the gap.

Then it clicked.

Excavating an offer out of your own content is the same repeatable process every single time.

→ Find the signal…replies and DMs, not likes.
→ Find the question you keep answering for free.
→ Pull the language in their words, not yours.
→ Turn that one insight into a few different shapes…coaching, done-for-you, a product.

Attach a way to actually test it before you build anything.

Anything that’s a repeatable process can be run by a machine.

So I built one.

Meet AI Offer Crafter.

I pointed it at my own archive…published issues…and let it run.

Here’s what came back, unattended, in one pass:

✅ Ingested published issues across the archive

✅ Generated 3 offer directions with confidence score with reasoning for each direction

✅ The output for each Offer Direction:

• What is included in the product

• The delivery reliability (time needed to deliver)

• How to actually build this offer (Build Path)

• Pricing with reasoning.

• Hand riser email (for validation)

• DM Scripts (for validation)

• The Offer Landing Page Outline

• An actual Validation Plan before building the actual product.

No week of re-reading my own archive.
No guessing at a price with no anchor.

And because it’s a system, I can run it again every ten posts, point it at a completely different niche, or run it for a client’s archive instead of mine.

Here’s what’s actually happening under the hood.

I’m not going to show you the literal prompts…those are the actual moat, and handing them over would just let someone build a worse copy of this in a weekend.

But here’s the architecture, and why it’s split the way it is.

Three stages:

1. Distill

Every source…every post, thread, and issue you paste in…gets read separately, in parallel, and reduced down to structured signal: themes, pain points, the audience’s own language. This is what makes 40+ posts tractable instead of overwhelming. You’re not asking one model to hold your whole archive in its head at once…you’re asking it to extract the useful part from each piece, one at a time.

You don’t read the haystack. It hands you the needle from every piece of hay, separately.

2. Synthesize

All of that extracted signal gets combined…into the repeated question, a bank of your audience’s actual phrases, and three distinct offer shapes. This is where signal over applause actually happens: replies and DMs get weighted heavily, likes get ignored entirely. A post with 3 replies asking “how do I do this myself” beats a post with 400 likes and silence in the comments, every time.

This is the stage that decides what you should build…not just what performed.

3. Package

The direction with the strongest signal gets expanded into the full deliverable: positioning statement, pricing band, what’s actually included, a build path, and outreach copy…a hand-raiser email and DM scripts, ready to send.

Three stages. One run. An offer, not just a score.

Why this isn’t the same as pasting your archive into ChatGPT

I get why that’s the instinct. Here’s the difference in practice.

Paste your content into an AI prompt, and you get something like this back:

“Consider offering a coaching program, a digital course, or a done-for-you service based on your expertise. Price it competitively and market it across your channels for maximum reach.”

Technically true. Completely useless. It could apply to literally anyone.

Here’s what AI Offer Crafter returned instead, from a real run:

“Your audience isn’t asking ‘how do I grow on Substack’ — they’re asking ‘how do I turn 40 subscribers into people who’ll pay me.’ That’s the offer: a paid audit of their first funnel, priced at $750, built for creators at your exact stage.”

Not only this… It completely understands how to build the offer step by step, the landing page outline, and the copy of the landing page, and also, it will provide you an actual plan to validate the offer. (Never build any without validation).

Same input. Completely different output. One’s a suggestion. The other’s an offer you could send to someone today.

I'm not the only one who thinks so.

Here’s what you get.

AI Offer Crafter is $29, once. Not a subscription…nothing to cancel, no monthly fee eating into the same budget the tool is trying to help you get out from under.

13 people have already grabbed it. It’s capped at the first 50 at this price…after that, the price goes up.

You run it on your own Anthropic API key…which sounds like a hassle until you see the actual cost: $0.30–$0.80 for a full run across ~15 sources.

You see the estimate before you run it, and the real number after. You’re paying Anthropic directly for exactly what you use…not paying me a markup on top of API calls I’d otherwise be marking up.

And it’s backed by this: if you don’t get 3 qualified hand-raiser replies within 14 days, I’ll rework the positioning with you at no charge. Not a refund policy. A promise to fix it until it works.

Get AI Offer Crafter 🚀

What’s coming next…

This is day one.

The guarantee you just read? Shipped this week. Here’s what’s next:

→ Receipts:
Every pattern it finds will link straight back to the actual posts it came from. No more taking the tool’s word for it.

→ Input strength meter:
Tells you in real time if you’ve given it enough to work with, before you waste a run.

→ Tiered pricing:
One price band becomes a full good/better/best ladder, automatically.

→ Email Sequences Builder:
Once you choose the offer direction, it will generate the email sequence for the launch and pre-launch series

→ The Offer Page Visualizer:
With your offer direction, the app will show you the landing page design with an actual copy of the landing page.

I’m building this in public…Buy in at $29 now, and every one of these ships to you free…no upsell, no “pro tier” later.

If you’re not ready to buy yet, that’s fine. Just hit reply and tell me what your repeated question turned out to be…the thing you keep explaining for free. I read every one, and it’s genuinely useful to me to see the pattern across all of you, not just my own archive.

Until next time,
Sharyph

A step-by-step system for creators at every stage…whether you don’t have an offer, have one that’s dead, or have one that’s almost working.

TL;DR: To build an offer that converts, mine your own published content for the question people keep asking you for free, then package that answer into a named offer with a specific outcome, proof, and risk reversal. Choose a service, coaching, or digital product based on audience size and how fast you need revenue.

A step-by-step system for creators at every stage…whether you don’t have an offer, have one that’s dead, or have one that’s almost working.

Let me guess how your week went.

You published. Maybe a newsletter issue. Maybe three LinkedIn posts. Maybe a Substack essay you were genuinely proud of. People read it. Some replied.

A few said “this was great.” And your revenue moved by exactly zero dollars.

So you did what every creator does. You assumed the problem was reach. Not enough eyeballs. Need to post more. Need to grow the list. Need to crack the algorithm.

Here’s what I’ve learned after reviewing 47 creator offers and making 700+ product sales without a single ad: you almost never have a traffic problem. You have an offer problem.

And the specific version of that problem is worse than most creators think, because it’s invisible. Your offer isn’t bad. It’s unclear. Unclear offers convert below 1% no matter how good the thing behind them actually is.

I’ve watched the same product go from 0.8% conversion to 4.2% conversion with zero changes to the content…same modules, same price, just different positioning. Same product. 5x the revenue.

And here’s the part that matters most: if you’ve been publishing for more than three months, you have already written your offer. You just haven’t extracted it yet. Most of this issue is about how to dig it out.

Let’s get into it. 👇

Outline:

Which Stage Are You At?

What Is an Offer, Actually?

Why Don’t Offers Convert?

How Do You Find Your Offer in Content You’ve Already Published?

Should You Sell a Service, Coaching, or a Digital Product?

How Do You Price an Offer Without Flinching?

How Do You Actually Sell It?

What’s the 7-Day Offer Sprint?

Before we begin, this post is sponsored by Outrank.

Outrank went from $300 to $300k MRR in 15 months, largely by running its own SEO playbook on itself, making it the most recommended SEO tool by ChatGPT & Claude.

It writes and publishes articles to your site daily, builds real backlinks through a DR-matched exchange network, and gets your brand cited in AI answers.

You connect your site once, and it keeps shipping while you build.

If SEO has been on your “later” list forever, this is the version that runs without you.

Get Started for FREE 🚀

Back to post...

Which Stage Are You At?

Most creators fall into one of three stages: no offer yet, an offer that exists but isn’t selling, or an offer you understand but still can’t convert. Stage 3 is the hardest, because the obvious fixes are already done. The real fix is usually rebuilding the offer from outside evidence, not inside assumptions.

Before the playbook, find yourself on this map. Your starting point determines which section you actually need.

Most readers of this newsletter are Stage 3. That’s the hardest place to be, because the obvious fixes are already done…and the real problem is usually that the offer was built from the inside out instead of the outside in.

We’ll fix that.

What Is an Offer, Actually?

An offer isn’t your product or service…it’s the package around it: a specific dream outcome, a named deliverable, a unique mechanism, proof, risk reversal, and a price anchored to the transformation. A product description lists features. An offer makes the buyer picture the result and removes their reason to hesitate.

(Stage 1: this section is for you. Stage 2 and 3 — skim and move on.)

Most creators use “product,” “service,” and “offer” interchangeably. They’re not the same thing, and the confusion is expensive. Your product or service is what you deliver. Your offer is how you package and frame it. Your lead magnet is free and exists to grow your list…not to make money.

Alex Hormozi defines an offer as what a business provides in exchange for money…and he calls it the foundational starting point for every transaction. Not the marketing. Not the funnel. The offer.

Here’s why this distinction decides whether you earn anything. Two creators sell the same 4-hour course on newsletter growth.

Creator A sells a product: “Newsletter Growth Course…video lessons, templates, community access. $197.”

Creator B sells an offer: “The 90-Day Newsletter System…go from 0 to 1,000 engaged subscribers using the 3-template system that grew 47 newsletters past their first 1K. Without paid ads or daily posting. $297.”

Same content. Creator B charges more and converts better.

Why? Because an offer contains seven things a product doesn’t:

• A specific dream outcome — “1,000 engaged subscribers,” not “grow your audience”

• A named deliverable — the thing has a name, not a category

• A unique mechanism — the “3-template system,” not “proven strategies”

• Proof — 47 newsletters, not “trusted by creators”

• Risk reversal — a guarantee that moves risk off the buyer

• A price anchored to the outcome — not to your hours or the page count

• Explicit exclusion — who this is not for

The three-sentence test: right now, without opening your website, write down who your offer is for, the specific measurable outcome, and the price. If you can’t do it in three sentences, that’s your problem. Not reach.

💡 If you’re at this stage and don’t yet know what you’d sell, run the Skill Monetization Matrix from The Claude Roadmap: From First Post to Full-Time Creator before you go further. It surfaces which of your skills are actually monetizable.

Subscribe now

Why Don’t Offers Convert?

Offers fail to convert for six reasons: vague outcomes, messaging pitched to the wrong awareness stage, no unique mechanism, no risk reversal, too many options, or selling to an audience that can’t afford the transformation. In 47 reviewed offers, 83% failed on clarity…not on the quality of what was being sold.

(Stage 2 and 3: this is your diagnostic. Run every one of these against your live offer.)

When I reviewed those 47 offers, 39 of them couldn’t answer “who is this for, and what exact transformation do they get?” That’s 83%….and every one of those creators had genuine expertise. The failure was never the expertise. Here are the six failure modes, in the order they actually kill offers.

1. Vague outcome

“Better content.” “Grow your business.” “Level up your writing.” None of these are outcomes. They’re categories. A buyer can’t picture themselves having achieved them, so they can’t want them.

Fix: Attach a number and a timeframe. “2 more clients per month.” “1,000 subscribers in 90 days.” “15 hours back every week.”

2. Wrong awareness stage

This one is subtle and it kills good offers. Eugene Schwartz mapped five stages of buyer awareness:

Unaware → Problem-Aware → Solution-Aware → Product-Aware → Most Aware.

Your messaging has to match where your reader actually is.

Most creator audiences arrive Problem-Aware (”my content isn’t making money”) or Solution-Aware (”I know I need an offer”).

But most creator sales pages are written for the Most Aware…they open with features and pricing, as if the reader has already decided.

Fix: If your audience is Problem-Aware, your page opens with the problem in their words, not your solution’s name.

3. No unique mechanism

Schwartz’s second model matters even more in 2026: market sophistication. In a fresh market, you can just state a claim. In an exhausted market, a plain claim is invisible. “I help creators grow on LinkedIn” is a level-5-sophistication claim being made at level 1. There are ten thousand of those. Your reader has scrolled past all of them.

Fix: Lead with the mechanism, not the promise. Not “grow on LinkedIn” — “the 3-template system.” Not “get more clients” — “the reply-mining method.”

💡 I went deeper on why copy-paste positioning stopped working in Copying Big Creators Won’t Work in 2026.

4. No risk reversal

Your buyer has been burned. They’ve bought courses they didn’t finish and hired freelancers who ghosted. Every purchase carries a silent question: what if this doesn’t work? If you don’t answer it, they answer it themselves. And the answer is always no.

Fix: A specific guarantee beats a vague one. Not “satisfaction guaranteed” — “if you don’t have your first paying client in 60 days, I’ll work with you free until you do.” The counter-argument (it attracts bad buyers) is real but rare; in practice, strong guarantees lift conversion more than they lift refunds…if you actually deliver.

5. Too many options

Three tiers, five add-ons, a payment plan, and a “custom” option. You think you’re serving everyone. You’re creating paralysis.

Fix: One clear path. If you must tier, use three and make the middle one obviously correct.

6. Selling to people who can’t pay

Hormozi’s bluntest rule: if your market can’t afford the transformation, no offer will save you. A brilliant offer sold to broke beginners loses to a mediocre offer sold to funded operators.

Fix: Check who’s actually replying to your content. If it’s students and hobbyists, your $2,000 offer isn’t broken…your audience is wrong for it.

💡 If you want to score your offer against these systematically, I built a Claude Skill that rates any offer across five dimensions in 60 seconds: I Built an AI Tool That Tells Me Why My Offers Don’t Sell.

How Do You Find Your Offer in Content You’ve Already Published?

Find your offer by auditing your published content for signal, not applause: sort posts by replies, DMs, and saves instead of likes. The question people keep asking you for free, in their own words, is your offer. Draft it three ways…done-with-you, done-for-you, do-it-yourself…then validate with a hand-raiser email before building anything.

(This is the main event. If you’re Stage 3, this is the section that changes things.)

Here’s the mistake almost every Stage 3 creator makes: they build the offer from the inside out. They ask “what am I good at? What could I teach? What would be impressive to sell?” Then they build it, launch it, and wait.

The creators who convert build from the outside in. They ask a completely different question: “What do people keep asking me for…in their words, not mine?”

The answer is already sitting in your published content, your replies, your DMs, and your comments. You’ve been running unpaid market research for months without reading the results. This is the method. Five steps.

Step 1 — Audit for signal, not applause

Open your archive. Newsletter issues, LinkedIn posts, Substack essays, YouTube videos, X posts. Everything from the last 6–12 months. Now sort them…not by likes or views.

Sort by:

• Replies (someone typed a response)

• DMs it generated

• Saves and shares

• “How do I do this?” comments

Likes are a reflex. A reply is a decision. A DM is a decision with friction attached. That friction is exactly what makes it predictive.

Creator-pricing analysts consistently report that saves and shares sit closest to actual purchase intent, while follower count is a lagging signal. Chenell Basilio, who reverse-engineers newsletter growth for a living, frames valuable content as content that optimizes for replies, DMs, shares, and saves…not passive likes.

Honest caveat: no controlled study proves reply → purchase for info products. But every documented case I’ve found runs through this door, so treat it as the strongest heuristic available. Your output: your top 10 posts by signal. Not by reach.

💡 Practical shortcut: export your archive into a sheet, drop it into Claude, and prompt: “Group these posts by underlying topic. Flag the clusters with the highest reply and save rates. Show me the exact phrases readers used in the replies.” If you haven’t set up a proper context file for this kind of work yet, start by building a dedicated business context file for Claude.

Step 2 — Find the question you keep answering for free

This is the whole ballgame. Somewhere in your DMs and replies, the same question keeps appearing in slightly different wording. You answer it every time. You’ve probably answered it fifty times. It feels too obvious to sell.

That obviousness is the tell. It’s obvious to you because you’ve solved it. It’s not obvious to them, which is why they keep asking.

Here’s the canonical case. Justin Welsh….now a $10M+ one-person business…found his first product by rereading his LinkedIn DMs. He was in SaaS sales, expecting sales questions. Instead his inbox was jammed with people asking how he’d grown his LinkedIn following.

He built a $50 course. First month: $10,482. It grossed roughly $75K over 15 months. Then he did the thing most creators never do…he treated that $50 product as what he called a “trust tripwire,” rebuilt it properly, and relaunched at ~$150.

First month of the new version: $94,651. And it converted 2.7x better than the $50 version. Higher price. Better conversion. Same underlying knowledge. His summary: your customers will tell you what to build next…you just have to listen.

Your output: the one question, written in your audience’s exact words. Not your paraphrase. Copy-paste the actual sentence someone sent you.

Step 3 — Extract their language, not yours

Amy Hoy and Alex Hillman named this “Sales Safari”…instead of asking your audience what they want, you observe them and take notes in four buckets:

• Pain — what they complain about

• Jargon — the exact words they use

• Worldview — what they believe about the problem

• Recommendations — what they’re currently buying instead

Watch for trigger words: “finally,” “easier,” “I hate when,” “I keep trying to,” “I gave up on.” This is what closes the gap between “I help creators with content systems” (your words) and “I stop you rewriting the same post for four platforms every Sunday” (their words).

Your offer copy should be roughly 70% their vocabulary and 30% yours. Your output: 15–20 verbatim phrases from your own audience.

Step 4 — Draft three offer shapes from the same insight

Do not draft one offer. Draft three, from the same core insight, at three different levels of involvement:

Same insight, three vehicles. This matters because Stage 3 creators usually picked the wrong vehicle, not the wrong topic. The topic was right. Turning it into a $97 course when your audience needed it done for them is why nobody bought.

Your output: three one-paragraph offer descriptions. Each with a name, an outcome with a number, and a mechanism.

Step 5 — Validate with a hand-raiser before you build anything

This is the step almost everyone skips, and it’s the cheapest thing in this entire playbook.

Send one email to your list. Not a pitch..a question. Something like: “I’m putting together something for people stuck on [exact problem in their words]. If that’s you, reply with the word MAP and I’ll send details.”

Then…and this is the part that matters…get on a call or a DM thread with everyone who replies. Ask what they’ve already tried. Ask what they’d pay to never deal with it again. Take notes in their words.

Chenell Basilio documented a version of this in her Growth in Reverse series: Nathan May, working with Jesse Pujji, emailed roughly 300 past readers asking them to reply if they wanted help…then ran 1:1 follow-ups to extract the pain in the readers’ own language.

That single hand-raiser email produced over $20,000, with a $47 low-ticket product laddering into multiple $5,000 upsells.

The insight from that case is exactly the thesis of this issue: the product-market-message alignment came from readers, not assumptions.

The rule that saves you months:

❌ Likes are not validation.
❌ “I’d totally buy that!” is not validation.

✅ Money is validation.
✅ A calendar booking is validation.
✅ “Send me the link right now” is validation.

Ask yourself after every test: did I get commitment, or did I get compliments? If fewer than a handful of people raise their hand from an engaged list, don’t build. Go back to Step 2…you picked the wrong question.

Should You Sell a Service, Coaching, or a Digital Product?

Pick service, coaching, or product based on your runway. Need cash in 30 days with under 1,000 people on your list? Sell a service. Have proven repeat demand and real traffic? Build a product. Fewer buyers at a higher price beats many buyers at a low price…a $2,000 service needs 3 clients; a $47 product needs 852.

You’ve got your insight and three possible shapes.

Now pick one…based on your constraints, not on what looks impressive on a landing page.

Run this top to bottom:

The uncomfortable math behind this tree:

Kevin Kelly’s 1,000 True Fans said 1,000 fans × $100/year = $100K.

Li Jin’s 100 True Fans revised it: 100 fans × $1,000/year gets you to the same place.

The critique of both is fair…converting a casual audience into true fans runs around 1%, and most creators never manage the direct relationship the model assumes.

But the practical lesson is the one that should drive your decision:
Fewer buyers at a higher price is dramatically easier than many buyers at a low price. If your list is 800 people, a $2,000 service needs 3 clients.

A $47 product needs 852 buyers…more than your entire list. One of those is achievable this month.

And the conversion math nobody wants to publish:
Substack’s official guidance suggests 5–10% free-to-paid.

That’s what winners do. Beehiiv’s State of Paid Newsletters 2026, drawing on first-party data across thousands of publications, puts the median at 0.62%…roughly 6 paying subscribers per 1,000.

Only about a fifth of publications are clear 5%. Plan your revenue at 1–3%. If you’re modeling 10%, you’re modeling a fantasy.

The progression almost everyone should follow…and the one Nathan Barry codified as the Ladders of Wealth Creation — is:

You can jump rungs. You can’t skip the skills. Selling a $497 course is a distribution problem you’re not equipped to solve until you’ve solved the delivery problem at $2,000-a-client scale.

💡 Once you’re ready to climb, the full rung-by-rung breakdown with real creator numbers is in The Creator Revenue Playbook.

One more thing about 2026 specifically

The market has split, and it matters for which offer you choose.

Commodity work…basic writing, basic design, simple builds…is getting compressed hard. Meanwhile, specialized implementation work is surging.

Fiverr’s Business Trends Index reported that searches for AI-agent freelance expertise jumped 18,347% over a six-month window, with workflow-automation and newsletter-setup specialists posting similar spikes. Fiverr’s own quarterly commentary put it plainly: specialized talent is outperforming while commoditized skills face pressure.

Translation for your offer: stop selling a deliverable. Start selling an outcome you own.

“I’ll write your newsletter” is a deliverable, and it’s being priced down.

“I’ll build the system that publishes your newsletter and repurposes it across three platforms” is an outcome, and it isn’t.

💡 If that direction fits you, I mapped 50 concrete versions of it in 50 AI Agents You Can Build and Sell.

How Do You Price an Offer Without Flinching?

Price the transformation, not your time or hours worked. If quoting your price doesn’t make you slightly nervous, it’s too low. Gumroad’s data across 146,000+ products shows pricing at $30 or higher correlates with roughly 9x more revenue per product than pricing under $10….underpricing doesn’t reduce risk, it attracts buyers who never implement.

Short section, because the principles are simple and the execution is emotional.

Price the transformation, not your time. Jonathan Stark and Blair Enns have spent careers on this single point: hourly billing caps your income at your calendar and punishes you for getting faster. What the client gets is what you charge for.

Use the discomfort test. If quoting the price doesn’t make you slightly nervous, it’s too low.

Don’t underprice to reduce risk. It doesn’t work. Underpricing attracts buyers who don’t value the outcome enough to implement it…which produces no results, no testimonials, and no referrals.

Gumroad’s analysis across 146,000+ products found pricing at $30+ correlated with roughly 9x more revenue per product than pricing under $10.

Price bands to anchor against:

Use the speed bumps. Price just under psychological thresholds: $97, $197, $497, $997, $1,997.

How Do You Actually Sell It?

Sell through your newsletter by teaching the problem for 2-3 issues, then opening with a real, time-bound reason before pitching. In DMs, warm beats cold: 50 messages to people who already engage with your content outperform 1,000 cold pitches. Track revenue per conversation, not messages sent, and always end your offer page with who it’s NOT for.

You have an offer. Here’s how it reaches people without you feeling like a used-car salesman.

The newsletter path

Your list is your conversion engine…social distributes, email converts. But most creators do one of two broken things: never mention the offer, or blast a hard pitch out of nowhere.

The sequence that works:

1. Teach the problem (2–3 issues). Genuinely useful content about the pain your offer solves. Soft P.S. link at the bottom.

2. Open with a reason. Founding-member pricing. Limited cohort. Real scarcity — never invented.

3. Handle one objection per email. Not five in one email. One.

4. Close with a deadline. A date, not a vibe.

One idea per email. One CTA per email. Write the CTA around the outcome — “Show me the system,” not “Learn more.”

The DM path

Warm beats cold, decisively. 50 DMs to people who already engage with your content will outperform 1,000 cold messages…and won’t damage your reputation.

The structure that works:

Line 1 — Specific reference. Something they actually posted. Not “love your content.”
Line 2 — The observation. A gap or pattern you noticed, offered as insight, not criticism.
Line 3 — The question. Not the pitch. “Is that something you’re working on right now?”

That’s it. No link in the first message. You’re opening a conversation, not closing a sale.

Track revenue per conversation, not DMs sent. Ten conversations that go somewhere beat 200 that go nowhere.

💡 I automated most of my own version of this…the research, the personalization, the drafting…in How I Built My Subscriber Pitch Agent.

The offer page structure

In order, every time:

1. Outcome headline (specific, numbered)

2. The problem, in their language

3. The offer + named deliverables

4. Proof

5. Risk reversal

6. Price

7. Who this is NOT for

8. Single CTA

Your tagline

Use this as a clarity tool, not a slogan:

“I help [specific person] achieve [specific outcome] through [unique mechanism] without [the thing they dread].”

April Dunford…the sharpest thinker on positioning…argues against treating fill-in-the-blank statements as your actual strategy, and she’s right: the formula captures your assumptions, not market truth.

Her process runs deeper (competitive alternatives → unique attributes → value → who cares → market frame).

So use the formula externally for clarity.
Use Dunford’s process internally to make sure the clarity is pointed at something true.

What’s the 7-Day Offer Sprint?

The 7-day sprint: pull your top 10 posts by replies and saves, find the repeated question in your own words, extract verbatim phrases, draft three offer shapes, run the decision tree, send a hand-raiser email, then talk to everyone who replies. You’ll have a validated offer, or proof you asked the wrong question in a week.

Everything above, compressed into a week. No new tools. No new subscriptions.

Day 1 — Pull your top 10 posts by replies, DMs, and saves. Not views.

Day 2 — Find the repeated question. Write it in your audience’s exact words.

Day 3 — Extract 15–20 verbatim phrases from your replies and comments.

Day 4 — Draft three offer shapes: done-with-you, done-for-you, do-it-yourself.

Day 5 — Run the decision tree. Pick one. Name it. Attach a number to the outcome.

Day 6 — Send the hand-raiser email. One question. No pitch.

Day 7 — Talk to everyone who replied. Ask what they’d pay to make this go away.

At the end of the week you either have a validated offer and a short list of buyers...or you have proof you picked the wrong question, which you found out in seven days instead of three months.

Both outcomes are wins. Only one of them costs you a quarter.

The One Thing

If you implement nothing else from this issue:

Stop asking “what should I sell?” and start asking “what do people keep asking me for?”

The first question is a guess. The second one has an answer, and the answer is sitting in your inbox right now.

Your offer isn’t something you need to invent. It’s something you need to notice.

Frequently Asked Questions

What’s the difference between an offer and a product?

A product or service is what you deliver. An offer is how you package it: a specific outcome, a named deliverable, a unique mechanism, proof, risk reversal, and a price anchored to the transformation. Two creators can sell the identical course…the one with a real offer around it charges more and converts better.

How do I know what offer to build if I don’t know what to sell?

You already know. Sort your published content by replies, DMs, and saves rather than likes, then find the question people keep asking you for free. That repeated question, written in their exact words, is your offer. Most creators who’ve published for 3+ months have already written it without noticing.

How long should it take to validate an offer before building it?

One week. Days 1-3 audit your archive and extract the repeated question in your audience’s language. Days 4-5 draft three offer shapes and pick one using the decision tree. Days 6-7 send a hand-raiser email and talk to everyone who replies. You’ll have a validated offer or proof to pivot…either way, in 7 days.

Should I build a service, coaching program, or digital product first?

Match it to your runway, not your ambition. Need revenue in 30 days with under 1,000 people on your list…sell a service. Have 3-6 months and you have delivered the same thing 3+ times…build a productized service. Have 12 months and proven repeat demand…build a digital product. You can skip stages, not skills.

What’s a reasonable price to charge for a first offer?

Price the transformation, not your time. Gumroad’s data across 146,000+ products shows items priced at $30+ generate roughly 9x more revenue than those under $10. Use psychological price points just under round numbers…$97, $197, $497, $997…and if the price doesn’t make you slightly nervous, it’s probably too low.

Key Takeaways

• Diagnose before you build: 83% of the 47 offers reviewed failed on clarity, not quality…run the six-reason checklist against your live offer before touching the product itself.

• Mine your archive for the offer: sort your last 6-12 months of content by replies, DMs, and saves, and find the question you’ve answered “for free” more than a dozen times.

• Draft three shapes, not one: turn the same insight into a done-with-you, done-for-you, and do-it-yourself version before picking a direction.

• Validate with money, not likes: send one hand-raiser email and treat only replies, bookings, or “send me the link” as real signal…plan to build only if a handful of an engaged list raises a hand.

• Match the vehicle to your runway: under 1,000 subscribers and need cash in 30 days, sell a service (2-3 clients); 12 months of runway and proven repeat demand, build the digital product.

Your Next Step

If you’re Stage 1 (publishing, no offer yet): Start with the free playbook…it covers niche, lead magnet, and getting your first content working before you spend a dollar. → Grab the Lean AI-Powered Business Playbook — Free

If you’re Stage 2 or 3 (offer exists, isn’t converting):
Do Day 1 and Day 2 of the sprint today. Then hit reply and send me your one repeated question…the thing people keep asking you. I read every reply, and I’ll tell you honestly whether there’s an offer in it.

That’s the whole ask this week. One reply. One question.

One more thing.

Everything in Part 3…the archive audit, the signal sorting, the language extraction, the three offer shapes…is a process. And processes that run the same way every time should eventually stop being manual.

Not just the offer: the tagline, the exact deliverables, the realistic time-to-deliver, how to find and approach the first clients, and the full build path for whichever direction you pick.

Same excavation method. Ten minutes instead of a week.

I’ll show you the whole build and how I use it to build offers and the complete offer system for my clients. (want to build your offer using this system..? just reply “offer”.)

Sharyph

P.S. If this issue clarified something for you, forward it to one creator who’s stuck publishing without earning. That’s the single fastest way this reaches the people who need it. And if you post about it, tag me on X @sharyph_ — I’ll jump in.

Until next time — Sharyph | Founder of The Digital Creator

Let me show you how I turned every new subscriber into a researched, qualified, personally-pitched sales lead…without writing a line of code

Last week I created this agent, and since you are here, here are some AI agents I have built before:

🚀 How I Built an AI Agent to Create Content for Your Brand.

🚀 How I Built an AI Agent to Research Content for Your Business

🚀 How I Built an AI Agent to Create Authoritative Social Media Content for Your Brand

And today, this is something very special for me…

Every time someone subscribes to your Substack, something happens. And then… nothing.

You get a notification. Maybe you glance at their profile if you’ve got a spare thirty seconds. Then you move on, because you’ve got fifteen other things pulling at you today, and “go research my new subscriber” isn’t one of them.

Here’s the part that stings: Some of those subscribers are exactly who you’d want to pitch. Founders. People who’ve raised a round. People with a live product, a landing page, real budget, real buying intent. They subscribed to you. They’re already warm. And you’ll never know it, because manually researching every new subscriber doesn’t scale past about subscriber number three.

So I stopped trying to do it manually.

I built an agent that catches the subscriber, researches them, decides if they’re worth pitching, writes the pitch in my own voice, sends it, and follows up if they go quiet..

I didn’t write a line of code. I didn’t even build a flowchart. I built the entire thing by talking to it, in plain English.

This week I’m showing you exactly what I built, how it works end to end, and what it’s already done in the wild.

Subscribe now

Meet the Subscriber Pitch Agent

Here’s the plain-English version:

Stage 1 — Catch the subscriber.

Substack sends a “New free subscriber to…” email to my Gmail every time someone joins. The agent watches for it, pulls the subscriber’s email and Substack profile link out of the email body, and logs it in a spreadsheet.

Stage 2 — Qualify the Person

Once a day, the agent goes back through anyone it hasn’t researched yet, visits their Substack profile, pulls their X, LinkedIn, and website links, and researches them..based my customer profile. If they clear the bar, they move to a “Prospects” list. If not, they’re just... a subscriber. No harm done.

Stage 3 — Draft the pitch, only when I say so.

This is the one place I stay in the loop on purpose. Nothing gets written up for send until I personally mark a prospect ready. When I do, the agent researches them further and drafts a fully personalized cold email…plus four follow-ups…using a “skill” I built specifically to write these pitches in my voice.

Stage 4 — Send it, once, cleanly.

Once I approve a draft, the agent sends the first email only (the follow-ups stay in reserve), grabs the email thread ID so it can track replies, and logs the send time.

Stage 5 — Follow up without me lifting a finger.

Every day, the agent checks if any sent prospect replied. If they did, it stops the sequence immediately…no awkward follow-up after someone already responded. If they didn’t, it sends the next follow-up on a set cadence, and gives up cleanly after five days.

That’s it.

So, why this…

This isn’t a “look what AI can do” toy. It’s a warm outbound channel that was sitting inside your subscriber list the whole time, doing nothing, because researching and personalizing outreach one person at a time is exactly the kind of task that eats your week and never gets prioritized.

You don’t need a sales team to have a lead qualification and outbound process. You need a system that runs the boring, repeatable steps…catch, research, qualify, draft, send, follow up…so the only decision left for you is the one that actually needs a human: is this pitch worth sending, yes or no.

Subscribe now

The Real output (full agent workflow)

1. New subscriber hit the inbox

2. Add a new row to the Google Sheet

3. Run the qualification process, based on my ICP, once done, I get an email like this: 👇

4. Updated the Prospect Sheet with Email Draft to approve

5. Approve by changing the status to “Ready to Send Email”

6. Send the email

7. Follow-ups

That’s the whole system.

Subscribe now

Now you may ask this:

Q: What if the agent can’t confidently tell if someone’s a good fit?
A: Build your qualification criteria to fail closed…if the research can’t verify a signal, it should not count in the prospect’s favor. It’s better to miss a borderline lead than to pitch someone based on a guess.

Q: Won’t this just spam people?
A: Only if you let it. The one deliberate gate in this build…Step 3…exists specifically so no email leaves your inbox without you personally reviewing the fit and approving the send. Automate the research, not the judgment call.

Q: What stops it from double-sending?
A: Two things…Step 3’s anti-loop check (won’t re-draft a row that’s already drafted) and Step 4’s skip condition (won’t send if ThreadID is already set). Build both checks in explicitly; don’t assume the platform handles it for you.

Let’s Build Step by Step:

Paid subscribers get the entire build, copy-paste ready:

• The exact prompt I used for every one of the five agents, word for word

• The exact Google Sheet column structure for both tabs

• The real gotcha I hit (a silent permissions failure) and the one-line fix

• How I built the pitch-writing “skill” itself…the actual instructions I gave it…plus a blueprint so you can build your own around any offer.

• The full, real output by showing what happens really when I run this. (Actual case)

• How to customize every stage for your own offer, audience, and qualification bar

If you’re building an audience and you’ve never turned a subscriber into a sales conversation without doing the research yourself, this is the exact system.

Subscribe now

Read more

Since you are here, here are some AI Agents I have built before:

🚀 How I Built an AI Agent to Create Content for Your Brand.

🚀 How I Built an AI Agent to Research Content for Your Business

🚀 How I Built an AI Agent to Create Authoritative Social Media Content for Your Brand

Subscribe now

How many hours have you spent clicking through your own website, page by page, checking whether the SEO is actually right?

If you’ve done it, you know the drill:

• Opening each page to count how long the title tag is

• Copy-pasting meta descriptions into a character counter

• Hunting for images with missing alt text

• Checking whether you accidentally have five H1s on the homepage

• Running one page at a time through PageSpeed

• Opening a competitor’s site in another tab and trying to eyeball what they’re doing that you’re not

You can lose an entire afternoon doing this for a 20-page site.

And here’s the part that actually stings: even after all that clicking, you still can’t answer the only question that matters…what should I fix first? You have a pile of observations and no priority order. You’re guessing.

I’ve been there.

For years I either paid someone $300–$500 for a one-off SEO audit, or I avoided the whole thing because I didn’t have a spare afternoon. The paid audits would land as a 40-page PDF, I’d skim it, feel overwhelmed, and file it away. Nothing changed. The money was spent, and the site stayed the same.

Then it clicked. An SEO audit is the same repeatable checks every single time. Title length. Meta length. H1 count. Alt-text coverage. Canonical present. Schema. Page speed. Compare to competitors. Rank by impact.

It’s a checklist…and anything that’s a checklist can be run by a machine.

So I built one.

Meet the SEO Audit Agent. I pointed it at a real website…gardendesign.com (sorry, guys if there is something on your website that you don’t want to disclose).

Here’s what came back, unattended, in one pass:

✅ Crawled and audited 24 pages at depth 2 (I can adjust this to 100+ pages and depth as well)

✅ Scored the entire site 94/100 (the garden design guys are doing a good job, I guess).

✅ Clustered every issue by severity into 5 medium and 2 low problem types

✅ Pulled a Lighthouse performance score for every page

✅ Benchmarked 2 competitors and surfaced the exact schema they have that this site is missing

✅ Produced a prioritized action plan, the fixes in order, with the pages attached

No afternoon lost. No $500 invoice.

And because it’s a system, I can run it again next week, or on a client’s site, or on ten client sites, for basically the same effort.

Here is the complete report. I want to see how detailed this is... (only for subscribers)

What you built

The SEO Audit Agent is one automation that takes a single URL and gives you back a complete, client-ready audit. You give it a website; it gives you a scored, prioritized report.

This isn’t theory. It’s a working system, and the output at the bottom of this issue is a real run.

At a high level, it’s built from four components:

1. The Crawler

Takes one seed URL and discovers the pages worth auditing — up to 25 of them at a crawl depth of 2, staying on the same domain. On gardendesign.com this found 24 pages automatically. You point at the front door; it maps the house. No manual URL list.

2. The Per-Page Analyzer

This is the engine. It scrapes each page’s raw HTML in a loop and pulls the signals that actually move rankings — title length, H1 count, word count, canonical, schema types, image alt-text coverage, internal links — then scores every page against a rubric. On the test run it graded each of the 24 pages from 87/100 (a B) to 98/100, with a one-line reason for every score.

3. The Performance Check

Runs each page through Google PageSpeed and captures the Lighthouse performance score plus Core Web Vitals. On the sample, mobile performance ranged from 58 to 93 — instantly showing which pages are dragging.

4. The Benchmark + Report Builder

Snapshots your competitors, compares their structure to yours, clusters all your issues by severity, and writes the final report with a prioritized action plan. This is the part that turns raw data into “here’s what to fix first.”

Four components. One run. A report you’d have paid for.

Read more

I was doing what most people do; opening a chat, typing out the same instructions I’d typed a hundred times before, copy-pasting old examples so the tone came out right, re-explaining my business every single time.

Then I learned there’s a way to stop starting from zero. It’s not a plugin you buy, not a subscription, not a secret model. It’s a plain text file. That’s it. A text file that tells Claude “here’s exactly how I do this job, do it this way every time.”

It’s called a Skill. And once I understood it, I rebuilt how my entire business runs on top of it.

By the end of this, here is what you are going to get:

• What Claude actually is, for anyone who’s never touched it

• What a Skill is (and the one-sentence idea that makes it click)

• What a Plugin is, and how Skills + Plugins work together

• Why this specific thing has creators obsessed right now

• The mistakes almost everyone makes when they start

• How to set one up for your business, correctly, in one sitting

• How this actually grows revenue…not just saves time

By the end, you won’t just understand Skills. You’ll have a folder of your own.

Before we start, throughout this post I will be using my Cluade Skill Kit for examples, so here is my Claude Skill Kit

Get it for FREE 🚀

WHAT CLAUDE IS, AND WHAT A SKILL ACTUALLY IS

Let’s back all the way up, because I want this guide to help you, even if you haven’t opened Claude before, and I don’t want to lose you in paragraph two.

Claude is an AI you talk to like a person. You type a message, it types back. No coding, no dashboard to learn. You could ask it to write an email, explain a contract, plan a trip, or draft your next newsletter…same way you’d ask a smart friend, except it never gets tired and it answers in seconds.

And..every new conversation begins from zero. Claude doesn’t remember that you run a newsletter, that you hate em-dash overload except when you use them on purpose, that your offer is $97 not $997. You either re-explain all of that every time, or you paste in a wall of context before you even ask your real question.

That’s the exact problem I had. And it’s the exact problem a Skill solves.

Here’s the one-sentence version:

Think about the difference between two hires:

• Hire A is talented, but it’s their first day. You explain your brand voice, your audience, your format, your rules…every single project.

• Hire B has a written playbook for the job. You say “write the newsletter” and they already know your voice, your structure, your do’s and don’ts. You just review and ship.

A Skill turns Claude into Hire B. Permanently, not just for one chat.

So what is a Skill, physically? It’s not an app, not a plugin you install from a store, not something you pay for. It’s a plain text file called SKILL.md. .md just means “Markdown”…a simple text format, nothing fancier than what you’re reading right now. Anyone can open it, read it, edit it in five minutes.

Every SKILL.md has two parts:

1. The description: a short note at the top telling Claude when to use this skill. Something like: “Use this whenever I ask for a newsletter, or say ‘draft this week’s issue.’”

2. The instructions: the actual playbook. The structure to follow, the tone rules, the things to never do, the exact format to hand back.

That’s it. That’s the whole mechanism. No code. No engineering degree required.

You’re not “building software”…you’re writing down how you already do the job, once, so Claude can do it the same way every time after.

One Skill fixes one job. Which is exactly why the next question matters: what happens when your business needs a dozen jobs done, in the right order, without you stitching them together by hand?

That’s where Plugins come in… (explained in next section)

Subscribe now

WHAT A PLUGIN IS, AND HOW SKILLS + PLUGINS WORK TOGETHER

One Skill fixes one job. Write a newsletter. Write a thread. Audit a tool stack. Each one is a specialist, sitting quietly until you need that exact task done.

But your business isn’t one job. It’s a sequence of jobs, in order, that add up to a result.

Find a topic → turn it into a hook → write the newsletter → chop it into a Substack Note and a LinkedIn post. Four jobs, four specialists, one outcome.

You could run each Skill one at a time, by hand, copying the output of one into the next. That works. It’s also exactly the kind of manual stitching-together we’re trying to get rid of.

Same hiring analogy from last section, one level up.

A Skill is one great employee with a clear job description.

A Plugin is a small department…several employees whose jobs are designed to hand off to each other in order, without you managing the handoff yourself.

Here’s what that actually looks like as a real chain, start to finish:

1. Find a topic — a Skill scans what’s trending in your niche right now and hands you a shortlist.

2. Write the hook — a Skill takes the topic you picked and writes five opening-line options built to earn the click.

3. Write the newsletter — a Skill takes the winning hook and writes the full issue around it, in your structure, your voice.

4. Repurpose it — a Skill takes that finished issue and turns it into a thread, a LinkedIn post, and a short-form Note.

One idea in. A topic, a hook, a full issue, and three more pieces of content out…in one sitting, without you rewriting the brief four times.

That’s the difference that actually matters: a Skill saves you from re-explaining one job. A Plugin saves you from re-explaining the entire workflow.

A couple of things worth clearing up here, because I got both wrong at first:

• A Plugin isn’t a different kind of technology. It’s just Skills, organized. If you understand a Skill, you already understand 90% of a Plugin.

• You don’t have to build the chain from scratch every time. You write the sequence once…“do step 1, then 2, then 3, then 4”…and reuse that same instruction every time you want that outcome. That one instruction is the plugin, in practice, even before you ever package it as an official installable thing.

This is also where people’s eyes light up when I explain it…because it’s the first moment Claude stops feeling like “a chatbot I ask questions” and starts feeling like “a small team running parts of my business.”

Which is exactly why this next part matters: why this specific idea has creators so worked up right now, and what’s actually driving that.

WHY CREATORS ARE SUDDENLY OBSESSED WITH THIS…

If you’ve been anywhere near X or LinkedIn in the last few months, you’ve probably seen it: people posting screenshots of folders full of .md files like they just found a cheat code.

Some of it’s hype. Most of it isn’t. Here’s the actual reason.

For two years, AI tools sold you speed. Skills are the first thing selling you continuity.

Every AI tool before this promised the same thing: faster writing, faster research, faster whatever. Useful, but it never compounded.

You were fast today, and tomorrow you were starting over, re-teaching the same AI the same things, at the same speed as yesterday.

Speed without memory is just a treadmill.

A Skill is the first version of “teach it once, keep the result forever.”

You’re not paying for a faster brain. You’re building a permanent asset…a written playbook that gets more valuable every time you refine it, not a prompt you retype and then lose in your chat history.

That’s the real shift.

One of those disappears the moment you close the tab.
The other one runs your business next week without you.

There’s a second reason, and it’s the one that actually matters for your bank account, not just your workflow.

Most of the creators are not short on ideas, they’re short on a repeatable process for turning ideas into finished, on-brand output without reinventing your approach every single time.

A Skill is that repeatable process, written down.

And the third reason, the one nobody says out loud, is that this is genuinely one of the few “learn this now” moments where being early actually pays off. Not because the technology disappears if you wait. Because the compounding does.

Every week you run your Skills, they get sharper…you notice what to adjust, what examples to swap in, what to cut. The person who started three months ago isn’t just “further along.” Their system is better, because it’s been used and refined three months longer than someone starting today.

That’s the obsession, stripped of hype: it’s the first AI habit that compounds instead of resets.

None of that matters, though, if you set it up wrong on day one…and almost everyone does.

That’s next: the specific mistakes I made (and still see people making) when they first start with Skills.

THE MISTAKES ALMOST EVERYONE MAKES
(I MADE ALL 3)

Before I show you how to set one up, I want to save you three months of doing it wrong.

Because I did all three of these, in order, before it clicked.

Mistake #1: Downloading 20 Skills on day one and using none of them well.

The moment you see a folder of 25 Skills sitting there, the instinct is to install all of them immediately. It feels productive. It isn’t. You end up with 25 half-understood tools and zero habits…you don’t know which one to reach for, so you default back to the plain chat box, and the whole folder just sits there unused.

The fix is almost insultingly simple: pick the one Skill tied to the task you already do most, and use only that one for a week.

Not because the others aren’t good…because a Skill only compounds once it’s a habit, and you can’t build a habit around 25 things at once.

Mistake #2: Never replacing the placeholders, so everything sounds like it came from someone else’s business.

Every Skill worth using ships with fill-in-the-blanks: [Your Brand], [your audience], [your offer link]. The mistake is treating those as optional. People run the Skill straight out of the box, get a perfectly good newsletter, for a business that isn’t theirs. It reads on-brand for nobody, because it’s on nobody’s brand yet.

A Skill only sounds like you once you’ve told it who you are…once, in one file. Skip that step and you’ve just automated generic.

Mistake #3: Treating a Skill like a one-off prompt instead of a reusable asset.

This one’s subtle. People write a great instruction, get a great result, and then... don’t save it anywhere.

Next week, same task, they retype the whole thing from memory, worse than the first version. That’s not a Skill…that’s just a prompt you happened to write twice.

A Skill is only doing its job if it lives somewhere permanent and gets slightly better every time you use it…

a line added here, an example swapped there. The version of my newsletter Skill I use today is not the version I wrote three months ago. It’s been edited maybe fifteen times.

That’s not a flaw in the system, that’s the system working.

One more, smaller than the first three but worth naming:

Starting with a content Skill before you’ve set up your brand context Skill.

It’s like handing Hire B a flawless job description for “write the newsletter”…but never telling them your name, your audience, or your voice. Good structure, wrong voice.

Set up the brand-context file first; every other Skill gets sharper the moment you do.

Every one of these mistakes has the same root: treating a Skill like a clever trick instead of a piece of infrastructure. Tricks get used once. Infrastructure gets built once and used forever.

So…how do you actually avoid all four of these and set one up correctly, in one sitting?

That’s next, and it’s the part where you stop reading about this and actually have your own folder running.

HOW TO SET ONE UP CORRECTLY, IN ONE SITTING

Everything so far has been the “why.”

This is the “how”…the exact sequence, start to finish, so you’re not guessing.

Step 1: Start with a Skill someone already wrote. Don’t write one from scratch on day one.

This is the opposite of what feels productive, but it’s the right move.

Writing a great Skill from a blank page takes iteration…you write it, use it, notice what’s missing, fix it, repeat. Skip that whole learning curve by starting from a real, already-working file and adjusting it, instead of staring at an empty document.

Here you go 🚀

Step 2: Set up your brand file first, before any content Skill.

Remember the mistake from last section: good structure, wrong voice, if you skip this.

Open the brand-context file, and fill in the blanks, your name, what you do, who you help, how you sound, your real links and offers.

This one file is what every other Skill reads from, so it’s worth doing properly, once.

You can get this prompt inside Create Skill Kit

Step 3: Load it into Claude, two ways, pick based on how often you’ll use it.

Option A — just paste it. Zero setup, works right now.
Open a new chat with Claude, paste the entire Skill’s text in, then type your actual request underneath it. For example, you can say…

“Now use this to write me a newsletter about [topic].”

Claude reads the Skill and follows it for that conversation. This is the fastest way to try a Skill before committing to it.

Option B — install it properly, so it loads itself.
This takes 3 minutes and means you never paste it again:

1. Go to Settings → Capabilities, and make sure “Code execution and file creation” is turned on. Skills won’t appear without this.

   

2. Go to Customize → Skills.

3. Click the “+” button, then “+ Create skill.”

4. Select “Upload a skill.”

5. Upload the Skill as a ZIP file — the zip needs to contain the whole folder (not just the loose SKILL.md file on its own).

6. It’ll appear in your skills list. Make sure the toggle next to it is switched on.

That’s it. From here, you don’t open the file again. You just ask for the task in plain language…“write this week’s newsletter”…and Claude recognizes it matches the Skill’s description and uses it automatically.

Step 4: Test it before you trust it.

Ask for the exact task the Skill is meant to handle, using natural language, not special commands.

If the output already sounds close to right, it’s working. If it feels generic, go back to Step 2; it’s almost always a placeholder you forgot to fill in.

HOW THIS ACTUALLY GROWS REVENUE, NOT JUST SAVES TIME

Everything up to here could sound like “a tidier way to use AI.” That undersells it.

Let me show you where this actually hits the business, not just the to-do list.

Every business…yours included…runs on four kinds of work, whether you’ve named them or not:

1. Content — the writing, the ideas, the stuff that earns attention.
2. Strategy — knowing what to make and when, instead of guessing.
3. Monetization — turning attention into an actual offer people pay for.
4. Systems — the plumbing that runs it without you touching every piece by hand.

Most creators are fully staffed on none of these. You’re doing all four yourself, badly, at 11pm.

A folder of Skills is what it looks like to finally have someone…or something… covering each of those four jobs.

Here’s what that looks like concretely, one per area:

• Content: a Skill that writes your newsletter, another that turns it into a Notes, another that finds what’s worth writing about before you even open a blank page.

• Strategy: a Skill that builds a full content calendar, so “what do I post today” stops being a daily decision you make from scratch.

• Monetization: a Skill that pressure-tests a product idea before you spend a weekend building it, another that writes the sales page once you know it’s worth building.

• Systems: a Skill that audits your tool stack and tells you what to cut, another that designs the automation for whatever’s left.

But the real unlock isn’t any single Skill…it’s chaining them, the way I showed you before.

One Skill saves you an afternoon.

A chain of them turns one idea into a finished, sellable thing without you managing every handoff.

Here’s a real one:

→ an idea comes in
→ a Skill checks if it’s actually worth building
→ a Skill packages it into an offer with real pricing
→ a Skill writes the sales page
→ a Skill writes the launch emails.

Four Skills, one weekend, one new offer live… instead of the version where that idea sits in your notes app for two months because “productize this” felt like a whole project on its own.

That’s the actual shift: you stop being the bottleneck in your own business.

Not because you’re doing less…because the parts that used to require you to sit down and re-figure-it-out from scratch now have a playbook that already knows how.

This is why I stopped thinking of these as “AI prompts” a while ago.

A prompt helps you do a task once.
A Skill, and a chain of them…is closer to hiring your first employee, except the salary is whatever you’re already paying for Claude.

That’s everything I wanted to walk you through…what these are, why they matter, how to set one up right, and where it actually touches revenue.

Last thing: the one-sentence version of all of this, and where to get the actual folder so you don’t have to build it from scratch.

GRAB ALL MY SKILLS

If you remember nothing else from this issue, remember this:

Everything else…the plugins, the chains, the four business areas…is just that one idea, applied.

Three months ago I was re-explaining my business to Claude every single day. Now I have a folder that already knows it…my voice, my structure, my offers, my rules…and it gets a little sharper every week I use it.

I didn’t want to just tell you about this and let you rebuild it from scratch. So I didn’t.

I packaged the entire thing…all 25 of the actual Skills running The Digital Creator, the plugin chains from Section 3 and 7, and the exact prompts to make every one of them sound like you instead of me…into one free download.

No “here’s 3 of the 25 to tease you.” It’s the whole thing, free, because the goal was never to hide the system — it was to prove it works by handing it over.

Get the free Creator Skills Kit →

Inside, start exactly where I told you to ,open the brand-context file, fill it in once, pick one Skill, and use it for a week before you touch the rest.

Keep it simple,
Sharyph
Founder of The Digital Creator

Welcome to The Digital Creator, helping creators accelerate with AI by showing you how to start, create, grow, and scale faster.

Before we begin today’s actual content, I want to introduce My Operation Command Center. Right now I am working on content repurposing workflows, but it is way broader than just content repurposing.

TL;DR: Content repurposing means rebuilding one newsletter natively for each platform…not copy-pasting it. Cross-posting risks originality penalties only from app watermarks, duplicate files, or zero-effort aggregation. A five-platform system (LinkedIn, X, Substack Notes, Instagram, video) built on a brand voice file turns one pillar into five native posts in 30 minutes.

You write one great newsletter. Then it lives and dies in one inbox.

It goes out. Your subscribers read it (or don’t). And then it just... sits there.

Doing one job, when it could be doing five.

That’s the quiet tragedy of most creator content. Not that it isn’t good. It’s that your best thinking gets trapped in one format, for one audience, on one day - and then you go back to a blank page and start over.

Here’s the thing though - this isn’t another “post more” newsletter. I’m not telling you to be on six platforms, feeding six algorithms, chasing six different audiences.

I’m going to show you the exact system I use to take one newsletter and turn it into five platform-native pieces of content - without touching a keyboard for most of it, and without it reading like AI wrote it. Thirty minutes, start to finish.

But first, we need to kill a bad assumption. Because most creators aren’t stuck because they can’t write. They’re stuck because they think the problem is volume.

It isn’t.

Is “Post More” Actually the Answer?

No. Sprout Social’s 2025 analysis of nearly 3 billion posts found brands publishing less (9.5 posts/day, down from 2023) while engagement rose almost 20%. The real fix isn’t more content…it’s distribution: reaching the audience you already have, spread across 6-7 platforms a month, instead of trapping every idea in one inbox.

Here’s what nobody tells you: the brands winning right now are posting less, not more.

Sprout Social’s 2025 Content Benchmarks Report analyzed almost 3 billion messages across roughly 1 million public profiles. What they found should change how you think about your content calendar: brands published an average of 9.5 posts a day in 2024 - down slightly from 2023 - while inbound engagement rose almost 20%, from 70 engagements a day to 83.

Fewer posts. More engagement. Sprout’s own words on it: social platforms are reaching peak saturation, and posting for the sake of posting only creates noise.

So if “post more” isn’t the answer, what is?

It’s distribution. Specifically - reaching the audience you already have, wherever they actually are. Because DataReportal’s Digital 2025 report, built on GWI data, puts the average person on roughly 6–7 different social platforms a month. Your audience isn’t sitting in one inbox waiting for you. They’re scattered.

One newsletter reaches the people who open newsletters, on the day it lands. Everyone else - the people who live on LinkedIn, the people who scroll X at 11pm, the people who only ever see Substack Notes - never sees your best idea at all.

And here’s the part that actually hurts: this is the exact thing burning creators out.

Billion Dollar Boy, in partnership with Censuswide, surveyed 1,000 creators in mid-2025. 52% said they’d experienced burnout directly because of their career. 37% had actively considered leaving the profession. The #1 named cause? Creative fatigue - cited by 40%.

Not lack of ideas. Fatigue.

Because most creators think the fix for “my content isn’t reaching enough people” is make more content. So they sit down and manually rewrite the same idea five different ways, for five different platforms, every single week.

That’s not a content problem. That’s a distribution problem wearing a content problem’s clothes.

The real fix isn’t willpower. It’s a system - one that takes the newsletter you already wrote and turns it into everything else it should have been, without you doing the manual labor five times over.

That’s what the rest of this issue is going to build.

Does Cross-Posting Actually Get You Penalized?

Rarely, and only for three specific triggers: carrying another app’s watermark, uploading byte-identical duplicate files, or aggregating content with zero added value. Your own idea, rewritten in your own words for the platform it’s landing on, isn’t a violation…it’s normal content strategy. Even flagged accounts recover within Instagram’s rolling 30-day window.

Here’s the belief that’s quietly costing you the most reach: “If I post the same idea on multiple platforms, I’ll get penalized for it.”

So creators do nothing. One newsletter. One inbox. Forever.

I want to correct this properly, because the correction is actually good news.

Platforms do not punish you for posting your own idea in more than one place. What they punish is much narrower - and once you see the three actual triggers, you’ll realize you were never in danger to begin with.

❌ Trigger 1: Other apps’ watermarks. Instagram head Adam Mosseri said it plainly back in October 2024: what they try not to do is recommend Reels carrying another app’s logo - but your own logo is fine, don’t worry about it. TikTok’s own Originality Policy says the same thing in different words: repurposed content carrying a watermark or logo generally doesn’t count as original.

❌ Trigger 2: Byte-identical duplicate files. Back in April 2024, Instagram confirmed that if it finds two or more identical pieces of content, it only recommends the original. Same file, uploaded twice - that’s the problem. Not the idea living in two places.

❌ Trigger 3: Aggregation with nothing added. This is the one that just got bigger. As of April 30, 2026, Instagram expanded this originality enforcement from Reels to photos and carousels too. Their exact standard, and I want you to actually read this line: watermarks or credits do not count as making something original. You need real transformation - commentary, structure, your own framing - not a border slapped on someone else’s work.

Here’s what none of these three things are: your own idea, rewritten in your own words, formatted for the platform it’s landing on. That’s not a violation. That’s just... doing your job.

💡 Here’s the part I actually want you to sit with: I went looking for the scary stat you’ve probably seen - “cross-posting cuts your reach by 42%,” or the “72% penalty with a watermark.” I could not find a single study behind either number. It traces back to one SaaS blog, no data, no methodology, nothing. It’s not research. It’s a made-up number that got repeated so many times it started sounding true.

That’s the real lesson here. Half the fear keeping creators stuck to one platform isn’t even based on anything real.

What’s the Difference Between Cross-Posting, Repurposing, and Atomizing?

Cross-posting uploads the same file unchanged and risks dedup penalties. Repurposing rebuilds the same idea natively for each platform’s format and reaches new audiences. Atomizing breaks one pillar into multiple standalone pieces that compound reach over time. This is COPE, Create Once, Publish Everywhere…a modular content model NPR’s team coined in 2009.

This isn’t a new idea, and it isn’t an AI-era invention, and it is by Daniel Jacobson’s team at NPR, describing a modular, format-agnostic approach to content. Jacobson himself noted that COPE is agnostic as to whether you build the system or buy it - which is a useful thing to remember later in this issue, when we get to the build-vs-buy conversation.

Sixteen years old. This principle was true before AI existed. AI just made the execution of it cheap enough that you no longer have an excuse to skip it.

So here’s the shift I want you to make:

Stop asking “should I post this everywhere?”

Start asking “have I rebuilt this natively for where it’s landing - or did I just copy-paste it and hope?”

That one question is the difference between getting demoted and getting discovered.

How Do You Know If Your Newsletter Is Repurpose-Ready?

A repurpose-ready pillar has 3-7 distinct teachable points, 1-3 real stories or examples, 1-2 sourced data points, and at least one genuine stance, not just a summary of the space. If your newsletter is missing these, fix the pillar first. Atomizing thin content just spreads thin content across five platforms instead of one.

Before we get into the mechanics, I need to stop you for one second. Because everything I’m about to show you only works if the thing you’re repurposing is actually good.

You can’t atomize thin content into thirty useful pieces. You just get thirty pieces of thin content.

Garbage in, garbage out…except now the garbage is spread across five platforms instead of one.

So before you run a single newsletter through a repurposing system, run it through this test. A genuinely repurpose-ready pillar has:

✅ 3–7 distinct, teachable points - not one big idea stretched thin

✅ 1–3 real stories or examples - yours, a client’s, a case study

✅ 1–2 sourced data points - something with a number and a name attached

✅ At least one real stance - a “here’s what I actually believe” moment, not just a summary of the space

If your newsletter doesn’t have at least that, don’t automate it yet. Fix the pillar first. Everything downstream depends on it.

Here’s why this matters more than it sounds like it should: Buffer’s analysis of 2.1 million Instagram posts across 102,000 accounts found that posting 3–5 times a week more than doubles follower growth compared to 1–2 times. Push to 6–9 times a week and growth climbs to roughly 3.7×. Push past 10 and you’re at roughly 5.5× - but the returns are clearly diminishing by that point. You’re doing 2× the work for a fraction of the extra lift.

Read that data correctly and it tells you something important: the goal was never “post as much as possible.” It’s “post genuinely native content, consistently, without burning out producing it.” Five real posts beats twenty noisy ones - every single time.

That’s the whole premise of this issue. Not more content. The same strong idea, made cheap to distribute properly.

💡 If you’re not sure your last few issues would pass this test…that’s worth sitting with before you build anything.

And if you’re just getting your first pillar off the ground, this is exactly what The Lean AI-Powered Business Playbook for Creators walks you through…building a clear offer and a content system before you ever touch automation. It’s free.

How Do You Adapt One Idea for Each Platform?

Each platform has its own physics. LinkedIn rewards posts that survive the 210-character “see more” fold. X’s open-sourced algorithm now weighs reply-chain depth over likes and penalizes in-body links. Substack Notes now reaches non-followers by default. Instagram rewards a 1-2 second hook. Copy-pasted text ignores all of this…and reads as low quality.

Okay. Your pillar passed the test. Now here’s where most people still get it wrong - not because they cross-post, but because they don’t actually translate.

Copy-pasting the same words into five different boxes isn’t repurposing. Each platform has its own physics. Ignore them, and even a genuinely native idea reads like it doesn’t belong there - and the algorithm reads that scroll-away as “low quality,” penalty or not.

Here’s what actually changes, platform by platform.

🔗 LinkedIn AuthoredUp’s analysis of 372,126 posts (Sept 2025–Feb 2026) found posts between 1,301–2,500 characters earn roughly 27% higher engagement than anything under 400. But the number that actually decides whether anyone reads that far is the “see more” fold - roughly 210 characters on desktop, and closer to 140 on mobile, where most of LinkedIn’s traffic lives. Your first two lines aren’t an intro. They’re the entire pitch.

On links: the old “just put it in the first comment” advice has gotten more complicated in 2026 - LinkedIn’s algorithm now recognizes that pattern too, and can flag it as “bridge behavior.” The safer move now is to write a post that stands on its own, mention the resource naturally, and only add the link where it feels earned - not to lead with the link as the entire point of the post.

And here’s the stat that should genuinely change how you feel about LinkedIn: 80% of LinkedIn profiles post just once a week. Show up 3–5 times with something real, and you’re already ahead of the vast majority of the platform.

🔗 X This is the platform where the rulebook just changed, and most creators haven’t caught up. In January 2026, X open-sourced its entire recommendation algorithm - a Grok-powered transformer model that reads your post’s actual content to rank it, rather than relying on hand-built rules.

Practical upshot: hashtags are close to irrelevant for discovery now. The model doesn’t need them to understand what you wrote.

What the code confirms actually matters: conversation depth - reply chains, especially ones you respond to - is the strongest signal, ahead of likes.

Early velocity in the first 30–60 minutes matters enormously. And external links in the body remain a real reach cost - put them in a reply instead.

🔗 Substack Notes (your home base - treat this one differently) Substack changed this feed in the second half of 2025, and it’s the single biggest shift most creators haven’t adjusted for: the Notes feed no longer prioritizes people who already follow you. Most of what you see now comes from writers you’ve never followed - which means most of what you post can reach people who’ve never heard of you, without you doing anything extra.

Restacks are the amplifier here, not likes. A single restack from someone with real reach can outperform your last dozen Notes combined. And a Note doesn’t read like a headline - it reads like a text to a friend. Short. A little unfinished. Not a summary of the newsletter - a single thought pulled out of it.

Here’s the throughline across every platform above: the algorithm isn’t punishing you for showing up in five places. It’s punishing you for showing up the same way in all five. Native format is the entire unlock.

Which raises the obvious next question: who has time to rewrite one idea five different ways, correctly, every single week?

That’s exactly what we’re building next.

Why Most AI Repurposing Fails

Here’s a confession before I show you the system: I built the generic version first, and it was garbage.

Anyone can wire up an AI model and tell it “turn this newsletter into a tweet.” I know, because that’s basically what I did the first time I tried. And it was fine. Generic-fine. The kind of fine that reads perfectly correctly and says absolutely nothing. Completely forgettable.

Because here’s what nobody tells you about AI-repurposed content: nobody wants five more pieces of nowhere-content clogging up their feed.

And it turns out this isn’t just my experience - it’s the industry’s, at scale. Adobe’s 2026 Creators’ Toolkit Report, surveying more than 16,000 creators across eight countries with The Harris Poll, found that 87% of creators using creative AI say it’s genuinely accelerated their business or audience growth. That’s real. AI works.

But look at the second number: 57% say the AI output still needs moderate or extensive editing before it’s ready to publish. And 85% say the final creative decision should stay with the human, not the model. AI gets you to a draft fast. It doesn’t get you to done.

(Worth naming honestly: Adobe sells the tools in this story, so read their 87% the way you’d read any vendor’s own research - a real signal, but not a neutral one.)

Here’s why that editing gap exists, and it’s the same reason my first attempt was forgettable: the model wasn’t told anything about who I am. No voice. No rules. No banned words. Just “write like a professional.”

And “professional” is exactly what everyone’s AI output sounds like right now…which means it sounds like no one.

The hard part - the part that took the most iteration when I built this properly - wasn’t teaching the tool a voice. It was teaching it my voice. Not “a professional tone.” My tone. My rules. My banned words.

That’s what a brand.md file actually is. One markdown file that acts as Claude’s operating manual for being you.

I broke down the full nine-section structure - Identity, Mission, Audience, Voice & Tone Rules, Vocabulary, Writing Samples, Formatting Preferences, Offers & Links, Hard Rules - in a previous issue, and the Writing Samples section is the single highest-leverage piece of the whole file.

Here’s the line that made this click for me: without it, you’re not building a system. You’re re-briefing a freelancer who forgets everything the moment they clock out. Fifteen-plus AI chats a week, three minutes of re-explaining who you are each time - that’s close to an hour a week just teaching a tool your own identity. Every week. Forever.

The fix is embarrassingly simple once you see it: write it down once. Banned words like “delve,” “unlock,” “leverage,” “supercharge,” “in today’s fast-paced world” - gone. Real writing samples in your actual voice - in.

📖 If you haven’t built yours yet, this is the full walkthrough: The One File That Makes Claude Write Like You

Once that file exists, the AI isn’t guessing at your voice anymore. It’s reading your operating manual. And that’s what turns “generic-fine” into something that actually sounds like you wrote it - at 2am, in five different formats, without you touching a keyboard.

Which brings us to the system itself.

The 30-Minute Workflow

Here’s my honest framing before I show you this: this is my demonstrated workflow, not an industry benchmark.

I’m not claiming a study proved “30 minutes” as some universal number.

I’m telling you this is genuinely how long it takes me, end to end, and thirty minutes is more than enough time to do it properly.

👉 Minutes 0–2: Generate. Paste the pillar - the newsletter, the link, whatever the core piece is - into the voice-trained system. Hit go. Every platform-native draft comes back in one to two minutes.

👉 Minutes 2–15: Edit like a human. This is the part that actually matters, and it’s the part AI can’t do for you. One specific number or story dropped into each draft. The hook adjusted to the exact platform spec we walked through earlier - the LinkedIn fold, the X thread opener, the Substack Note that sounds like a text and not a summary.

👉 Minutes 15–25: Go native. Format touches. A carousel or clip where it fits. Links moved out of the body and into comments or replies, the way each platform actually wants them handled.

👉 Minutes 25–30: Stagger and schedule. Don’t drop all five at once. Space them out.

That’s it. Thirty minutes, and one newsletter has become five platform-native pieces of content instead of one thing sitting in one inbox.

Here’s exactly how I built the system that does the heavy lifting in that first two-minute step.

Repurposing HQ

I named it exactly what it does. You give it one newsletter, and it hands you back a full set of platform-native content….without you touching a keyboard for the repurposing part.

Paste a Substack link, or paste raw text if the link extraction fails…it basically never fails, because raw text is always the fallback. It pulls the article, cleans it, and rewrites the core idea per platform.

It’s built on three parts, and each one solves a different problem:

👉 The Reading Part - pulls and cleans the article from wherever it lives. Link or raw text, it doesn’t matter. This part just makes sure nothing gets lost going in.

👉 The Voice Part - this is the brand.md file we just talked about, doing its actual job. The core claim, and the one that matters most: it doesn’t sound like AI wrote it. It sounds like you did.

👉 The Writing Part - not one generic prompt for everything. Baked-in, per-format rules. X threads get a hook-first structure with a strict character limit per tweet. LinkedIn posts get built around the “see more” cutoff, because if your hook doesn’t land in the first couple hundred characters, nobody ever taps to expand it.

v2 output, formats:

The full build - architecture, prompts, the exact iteration that got the voice right - is broken down in a previous issue if you want to go deeper on how it actually works under the hood:

📖 I Built My Own AI Repurposing Tool. Here’s Exactly How.

And this is the whole mindset shift I want you to leave this section with:

Not “post more.” Not “hire a VA.” Just - take the thing you already wrote, and let a system that knows your voice turn it into everything else it should have been.

Should You Build or Buy a Repurposing Tool?

Build the layer you care most about…for most creators, that’s voice and text, since generic AI output costs you the most editing time there. Buy the commodity layers: video clipping, scheduling, anything already solved well and sold cheap. If a $15-29/month tool does the job, rent it instead of building it yourself.

Let me be upfront about my bias here, because you deserve to know it before I make the case: I built Repurposing HQ myself.

So take everything in this section with that in mind…and then let me argue against myself anyway, because the honest answer isn’t “always build.”

Here’s where I actually stand: more tools doesn’t mean more productive. My own tech stack used to be 21 tools costing $340 a month. Today it’s three - Claude, Kit, and Typefully - at $88 a month.

If I can’t explain how a tool directly generates revenue or saves me five-plus hours a month, it doesn’t make the cut. That’s the filter everything in my business runs through, and it’s the filter I want you to run this decision through too.

The case for building the voice layer:

Off-the-shelf AI repurposing tools give you “generic-fine” by default - because they don’t know your voice, only “a professional tone.” You end up editing just as much as if you’d written it yourself, except now you’re editing something that doesn’t sound like you to begin with.

There’s also the pricing shape of it. Most SaaS tools in this space run on a per-seat or per-credit model that scales against you - the more you use it, the more it costs, right when you need it most.

And there’s a quieter argument too: one integrated pipeline beats five disconnected tools stitched together with duct tape. You own the asset. You can extend it whenever you want, without waiting on someone else’s roadmap.

Even the person who coined the whole “Create Once, Publish Everywhere” model agrees this isn’t a build-only philosophy - Daniel Jacobson himself described COPE as agnostic to whether you build or buy.

So don’t mistake this section for me telling you to build everything from scratch. That’s not the argument.

The mature answer, and the one I actually live by:

✅ Build the layer you genuinely care about - for me, that’s voice and text, because that’s where “generic” actually costs me the most.

✅ Buy the commodity layers - video clipping, scheduling, anything that’s already been solved well by someone else and sold cheap.

If a $15–29/month tool already does the commodity job well, buy it. Don’t build what you can rent for less than an hour of your own time is worth.

Core + Secondary

Now I need to hand you the honest counterargument to everything in this issue - because if I don’t, this becomes hype, and hype is exactly what I promised you this newsletter isn’t.

Here’s the pushback, and it’s a fair one:

Post once a week across five different platforms, and every algorithm treats you as a casual, low-priority account on all five. You don’t build depth anywhere. You build a thin presence everywhere.

This is where the reconciling model actually matters - hub and spoke, sometimes called Core + Secondary.

One platform where the real relationship with your audience lives. Everything else repurposes outward from that core. You don’t chase presence on a platform you’ll quietly abandon in six weeks.

Look at how Justin Welsh actually built his - publicly, he calls it a Content Operating System: one newsletter, and 6–12 social pieces a week built outward from it.

His own site puts it at 500M+ yearly content impressions across LinkedIn and X, and an independent breakdown of his business puts revenue north of $12.5M since 2019, at roughly 86% margin.

That’s not five platforms run shallow. That’s one strong core, repurposed outward, relentlessly.

For you, the structure is already sitting there: the newsletter is your pillar. Substack is your home base.

Everything in this playbook - the translation rules, the workflow, Repurposing HQ - exists to take that core and extend it outward, not to replace it with five competing priorities.

A rule of thumb worth holding onto:

The honest contrarian take doesn’t undo this whole playbook. It defines the shape of it. Repurpose outward from something strong. Don’t chase presence everywhere at once.

Frequently Asked Questions

Will cross-posting the same newsletter idea get me penalized on social media?

No, not for posting the idea itself. Penalties only trigger from carrying another app’s watermark, uploading byte-identical duplicate files, or aggregating content with nothing added. A genuinely rewritten, platform-native version of your own idea has never been the problem…and even flagged accounts recover within a 30-day window.

How long does it actually take to repurpose one newsletter into five platforms?

About 30 minutes, split into four stages: 2 minutes to generate platform-native drafts from a voice-trained system, 13 minutes to edit in real specifics and adjust each hook, 10 minutes for native formatting touches, and 5 minutes to stagger publishing across the day instead of dropping everything at once.

Should I post on all five platforms if I’m just starting out?

No. Use the Core + Secondary model instead…one platform (your newsletter) as the pillar, one genuine secondary you’ll actually sustain. If you’re under 1,000 followers on a platform, treat it as an experiment, not a working spoke. Depth on two platforms beats a thin presence on five.

What should go in a brand voice file for AI content repurposing?

A brand.md file needs nine sections: Identity, Mission, Audience, Voice & Tone Rules, Vocabulary, Writing Samples, Formatting Preferences, Offers & Links, and Hard Rules. Writing Samples is the highest-leverage section…it’s what stops AI output from reading like generic “professional tone” instead of your actual voice.

We’ve come to an end. What’s the first platform you’re going to rebuild your last issue for?

Just reply to this email…I read every one.

Whenever you’re ready to take the next step:

→ The Lean AI-Powered Business Playbook for Creators - free. Start here if your pillar isn’t strong yet.

→ Premium Membership - $39/month or $150/year. Claude Skills, the AI Agents Collection, the Substack SEO/AEO toolkit, and 45% off every paid course.

→ Book a call if you are someone who wants a system like this to implement for your brand.

If you want to talk through where your content system actually stands right now, hop on a call with me - happy to help you figure out the next move.

Until next time - Sharyph | Founder of The Digital Creator

P.S. - If you build your brand.md this week, reply and tell me what your banned words list looks like. I read every one, and I’m always curious what “sounds like AI” to you specifically.

P.P.S. - Know a creator drowning in five platforms and zero system? Forward this one to them. That’s usually the moment it clicks.

Overview and my Mission

The Brand Command Canter will be the place where I do all the work, from generating ideas to creating content, repurposing, running campaigns, etc. And my mission with this system is to cut 80% of my repetitive work time and allocate that time for more creative works and building.

How to Set Up this System for your Brand

I have made it very easy to set this up. Below is the setup guide. Before you begin, the Complete system source code file is available for Paid Members only.

If you are a paid member, here is the link to download the file 👇

Subscribe now

Setup Guide

No coding experience needed. Just follow each step in order and copy/paste exactly what’s shown. It takes about 20-30 minutes.

Before you start, create two free accounts

• Supabase — supabase.com (this becomes your app’s database and login system)

• Anthropic — console.anthropic.com (this powers the AI — you’ll need to add a card, since it’s pay-per-use)

Step 1 — Install two free programs

• VS Code — code.visualstudio.com (where you’ll open the project)

• Claude Code — search “Claude Code” in VS Code’s Extensions panel (the icon of four squares on the left sidebar) and install it

Step 2 — Unzip the project and open it

Unzip the file you downloaded. It becomes one folder. Open VS Code, click

File → Open Folder, and select that folder.

Step 3 — Open the Terminal

In VS Code, click Terminal → New Terminal. A text box opens at the bottom — this is where you’ll paste a few commands. You won’t need to understand them, just copy, paste, and press Enter.

Paste this and press Enter:

npm install -g pnpm
pnpm install

(This downloads everything the app needs to run. It may take a minute or two.)

Step 4 — Create your Supabase project

1. Go to supabase.com/dashboard → New project.

2. Give it any name, and set a password — write it down, you’ll need it in Step 6.

3. Wait about 2 minutes while it sets up.

Step 5 — Create your settings file

Back in the VS Code Terminal, paste and press Enter:

cp .env.example .env.local

In the file list on the left, click on the new .env.local file to open it. You’ll paste your keys into it over the next two steps.

Step 6 — Add your Supabase keys

In your Supabase project, click Settings → API.

Copy each value below into the matching line in .env.local, then save the file (Cmd/Ctrl+S):

In Supabase, copy...Paste into this line in .env.localProject

Then click Settings → Database → Connection string:

Leave NEXT_PUBLIC_SITE_URL as it already is.

Step 7 — Add your Anthropic key

Go to console.anthropic.com → API Keys → Create Key.

Copy it into the ANTHROPIC_API_KEY line in .env.local. Save the file.

(Optional: for AI-illustrated carousel icons, get a free key at aistudio.google.com/apikey and paste it into GEMINI_API_KEY. Skip this if you don’t need it…everything else still works.)

Step 8 — One more Supabase setting

In Supabase, click Authentication → URL Configuration. Under Redirect URLs, add:

http://localhost:3000/auth/callback

Save.

Step 9 — Set up your database

Back in the VS Code Terminal:

pnpm db:migrate

This creates everything the app needs to store your data.

Step 10 — Start the app

pnpm dev

Open

http://localhost:3000

In your browser, sign up with any email and password, and you’re in. It’ll still show the original branding — that’s fixed in the next step.

Step 11 — Make it yours

In VS Code, click the Claude Code icon on the left sidebar to open it. Open the file docs/REBRAND_PROMPT.md, fill in your brand name, colors, and handle at the top of it, then copy the whole file and paste it as a message to Claude Code. It will update the app to match your brand.

When it’s done, go back to the Terminal, press Ctrl+C to stop the app, then run pnpm dev again to see your changes.

If something goes wrong

• Can’t sign in / stuck on “check your email” — in Supabase, go to Authentication → Providers → Email and turn off “Confirm email.”

• A command doesn’t work — make sure you’re in the right folder in the Terminal (it should show your project’s folder name).

• Anything else — open Claude Code and just describe what you’re seeing. It can read the actual error and fix it for you.

The System: Brand Command Center

Update 1:

The functions included:

Here are the main functions added for this version:

• Dashboard, where you can see an overview

• Create a brand profile (Brand voice)

• Create social media posts from newsletter/blog content

  • Substack Note

  • X Thread

  • LinkedIn Post

  • LinkedIn Newsletter

Subscribe now

Update 2:

What is included:

Here is what comes with this update.

• Dashboard view improved; you can see how many different types of posts scheduled

• Added a new feature called “calendar,” where you can adjust date and time for each post type

• Added a new feature called “LinkedIn Carousel,” where you can generate LinkedIn carousel from your content. This comes with dedicated post copy.

• Added a new feature called “Lead Magnet”, where you can create a PDF as a lead magnet from your content.

Subscribe now

This page will be updated once I push a new update, and if you are a paid member, you will be notified, and the full system source code will be shared with you.

That’s the quiet tragedy of most creator content. You spend hours getting a newsletter right, hit publish, and then... nothing. It doesn’t become a LinkedIn post. It doesn’t become Substack Notes. It just sits there, doing one job, when it could be doing five.

I know why. Because I did this for years.

Here’s the thing: the problem was never that I didn’t know repurposing was valuable. Every creator knows it. The problem is that doing it manually is boring, slow, and easy to skip.

You finish a newsletter, you’re mentally done, and the idea of opening five more tabs to reformat the same idea for five more platforms feels like homework nobody assigned you.

So most of it never happens.

Your best thinking gets trapped in one format, for one audience, on one day.

I got tired of that. So I built a tool to fix it.

What I actually built

I call it Repurposing HQ, and the idea is stupidly simple:

You paste in a link to your Substack post, or just paste the raw text if you’d rather. The tool reads it, pulls out the clean article, and then rewrites that same core idea into the formats that actually work on each platform.

✅ A short, punchy Substack Note.

✅ A LinkedIn post shaped the way LinkedIn posts need to be shaped.

✅ A Twitter thread that opens with a real hook instead of “so I wrote a new post.”

✅ A LinkedIn newsletter version, if that’s part of your world…

Right now it does four formats. That’s the first version I built. More are coming, and I’ll get to that.

But the part I actually care about more than the formats is this: it doesn’t sound like AI wrote it. It sounds like you did.

The Digital Creator

Accelerating Creators with AI Agents. I show how to use AI to start, create, grow, and scale faster.

By Sharyph

The part that actually mattered

Anyone can wire up an AI model and tell it “turn this newsletter into a tweet.” I know, because that’s basically what I did first. And it was fine. Generic-fine.

The kind of content that’s technically correct and completely forgettable, because it reads like it came from nowhere in particular.

That’s not useful. Nobody wants five more pieces of nowhere-content clogging up their feed.

So the real work…

Not “a professional tone.” Your tone. Your rules. Your banned words.
The stuff you’d never say and the stuff you always say.

Here’s how that actually works: you build a Brand Voice profile inside the tool.

Then, every single time it generates something, it looks yours… Not a hypothetical writer’s voice. Yours.

I tested this the hard way.

How it’s actually built (no jargon, promise)

I’m not going to turn this into a technical post…that’s what the video walkthrough is for. But here’s the shape of it, in plain English.

There are three moving parts.

Part 1: Reading Part

When you give it a Substack link, it has to go grab the actual article and clean it up… strip out the images, the formatting junk, all the stuff that isn’t the actual writing.

If a link doesn’t work for some reason, it falls back to just reading the raw text you paste in. That way it basically never fails on you.

Part 2: Voice Part

Everything I just walked through above. Your voice, your rules, stored and reused every time.

Part 3: Writing Part

Where the actual AI comes in, which is what I use for the writing itself. But here’s the important bit: it’s not one generic prompt for everything.

Each format…the Note, the thread, the LinkedIn post…has its own specific instructions baked in, based on what actually works on that platform.

Twitter threads get a hook-first structure with a strict character limit per tweet.

LinkedIn posts get built around the “see more” cutoff, because if your hook doesn’t land in the first couple hundred characters, nobody ever taps to expand it.

Little platform-specific rules like that, applied automatically, every time.

All of that lives inside a simple web app I built from scratch…

What this actually means for you

If you’re a creator sitting on a backlog of newsletters that never made it anywhere else, this is the fix.

Not “post more.” Not “hire a VA.” Just: take the thing you already wrote, and let a system that knows your voice turn it into everything else it should have been in the first place.

You don’t need my exact tool for this lesson to matter. Even if you never build anything, the mindset shift is the valuable part: stop thinking of repurposing as a manual chore you’ll get to eventually, and start thinking of it as a system problem. Systems get solved once.

Chores get skipped forever.

Subscribe now

What’s next

This is version one.

Four formats, working well, tested on my own content.

Coming next:

🚀 Long-form X articles,

🚀 LinkedIn carousels,

🚀 Infographic posts,

🚀 Lead-magnet generator that turns any newsletter into a downloadable PDF with its own promotional posts attached.

And because I know some of you don’t just want to hear about this…you want to see it…Below I have shared a full video walkthrough, showing you the actual system, how it’s built, and how you could build your own version of this, step by step.

Subscribe now

How I Build the Complete System

First let me show you the first build, what it does, and how I am going to use it.

Previous newsletter I mentioned in this video here: How to create a Brand .md file

The Complete Build

Now, let me walk you through the complete process of how I build this. Starting from researching. I have shared all, even the prompts, the struggles, the mistakes I made… above all what is my plan with this for the next version…👇

Read more

You publish. You refresh Google. You wait.

Meanwhile the person you actually wrote the article for isn’t typing into a search bar at all. They’re asking ChatGPT. Or Perplexity. Or Claude. And when they ask the exact question your post answers, the model either quotes you…or skips you.

An unoptimized post is now invisible twice.

The usual advice is “learn AEO.” Sure. So now it’s two checklists, two sets of rules, and one more job waiting for you at 11pm after the writing is already done. Which means most of us do it once, badly, and never again.

The wrong answer is a longer checklist. The right answer is a system that runs the checklist without you.

That’s why this guest post stopped me.

Gencay doesn’t optimize his articles anymore. He drops the finished draft into a folder called /finished and goes to sleep. Overnight, a Claude routine scores it 0–100 against SEO, AEO, and tone, rewrites the weakest sections, and scores it again…and it does not stop until it clears 85.

He reads the scorecard at breakfast.

Not a tool he bought. A skill he wrote. One folder.

Gencay writes Learn With Me AI — where he builds AI systems and Claude skills in public, and hands you the files.

Subscribe here → learnwithmeai.com

Below, he walks the whole build: the skill file itself, the first run on a real scheduled post (42/100), the rewrite loop (94 by round two), and how he turned it into a 7am routine that empties the folder while he sleeps.

One detail worth reading closely: on the rewrite, the loop lowered its own tone score — 20 down to 19 — rather than quietly editing his sentences to buy the point back. It flagged the line and handed it to him. A loop that polishes your voice to win its own game is worthless. This one took the worse number and said why.

If you write, publish, and then hope — this is the half of the job you’ve been skipping.

Here’s Gencay. 👇

Turning your idea into a finished article feels like the whole job.

It is half.

The other half is playing the distribution game by its rules. And the rules changed. Optimizing an article used to mean one thing: search engines. Now there is a new player on the field, AI.

Most of the people your article is written for ask their questions to AI now. ChatGPT, Perplexity, Claude. When someone asks the question your article answers, the AI either quotes you or skips you. An unoptimized post is invisible twice, once to Google, once to the machines your readers actually talk to.

But if AI is powerful enough to skip your article, it is powerful enough to fix it.

So I wrote a skill that does the heavy lifting. I finish the article, drop it into a folder called /finished, and go to sleep. In the morning, a Claude routine scores it from 0 to 100 against SEO, AEO, and tone rules. Below 85, it rewrites the weakest sections and scores again. It does not stop until the article reaches 85.

I read the report at breakfast.

The scoring is one skill. The 85 target is the goal. The retrying is the loop.

Here is the whole build.

The skill that scores, rewrites, and scores again

Read more

How many times last week did you paste the same context into Claude before it would write anything useful?

You know the ritual. New chat. Then the setup paragraph:

“I’m a creator in the AI space, my audience is solopreneurs, my tone is direct and no-fluff, don’t use corporate language, here’s an example of how I write…”

Then it finally produces something. Then tomorrow you open a new chat and do the whole thing again.

Here’s what that ritual actually costs you:

• Re-typing your positioning every single session

• Pasting the same three writing samples over and over

• Correcting the same mistakes (”stop using ‘delve,’ stop using ‘unlock,’ stop with the em-dash essays”)

• Getting output that sounds like every other AI creator because you never told it what makes you different

• Throwing away drafts that are 80% right but wrong in a way you can’t be bothered to fix again

And even after all that setup…the writing still drifts.

By message four it’s back to sounding like a LinkedIn robot. You tighten it, it drifts again. You’re not building a system. You’re re-briefing a freelancer who forgets everything the moment they clock out.

I did this for months. I told myself it was fine because each setup “only” took two or three minutes. Then I actually counted. Fifteen-plus chats a week, three minutes of re-briefing each…that’s close to an hour a week spent teaching Claude who I am. Every week. Forever. And I still wasn’t getting consistent output.

Then it clicked. The problem was never Claude. The problem was that I was keeping my brand in my head instead of in a file.

That’s why I built a brand.md.

It’s one markdown file. It holds everything Claude needs to write as me…my positioning, my audience, my voice rules, my banned words, real samples of how I actually write. I write it once. I load it once.

And now every surface where I use Claude…the chat app, Cowork, Claude Code…writes in my voice from the first message. No re-briefing. No drift by message four.

✅ Same voice across every tool

✅ Zero setup per chat

✅ New drafts that need editing, not rewriting

✅ A brand you can hand to a VA, a collaborator, or a future you

By the end of this issue you’ll have:

• The full anatomy of a brand.md — every section and why each one earns its place

• A from-scratch build guide, whether you have zero brand, a pile of old posts, or a clear brand with no file

• A complete copy-paste Prompt Template

• The exact way to deploy it in Claude Chat, Cowork, and Claude Code — because each one loads the file differently, and getting that wrong is why most people’s context files quietly do nothing

Let’s build it.

Meet the brand.md File

Not a brand deck. Not a 40-page guidelines PDF your designer made.

A lean, plain-text file…the kind you can read top to bottom in ninety seconds…that tells Claude who you are, who you’re talking to, and exactly how you sound.

Here’s the one mental model that makes all of this work. Every place you use Claude does one of two things with your file:

It either loads the file in full every time…or it retrieves pieces on demand.

That single distinction decides how you write the file and where you put it.

Claude Code loads it in full at the start of every session.

Claude Chat projects retrieve from it as needed.

Cowork reads it as a real file when you tell it to.

We’ll come back to this…but hold onto it, because it’s the thing 90% of people get wrong.

Why markdown, not a Google Doc or a Notion page?

Because markdown is plain text with light structure…headers, bullets, code blocks.

Claude parses that structure the same way you do: it reads the headers to know what’s what. No formatting to strip, no export step, no app to open. One file that lives anywhere and drops into everything.

This isn’t theory. This is the file I load before I write a single thread, newsletter, or sales page. Let’s break down what goes in it.

The Anatomy: What Goes In, and Why Each Part Earns Its Place

A brand.md has nine sections. Everyone is there for a reason. Skip one and Claude fills the gap with generic…which is exactly what you’re trying to escape.

1. Identity — who you are in one block

Your name, what you do, and your one-line positioning.

Why it matters: This anchors everything. Claude writes differently for “a solo creator teaching AI systems” than for “a SaaS marketing agency.” Without it, the model defaults to a bland average of the internet. This is the ground the whole voice stands on.

2. Mission & POV — what you actually believe

Your mission, plus your contrarian takes…the opinions that separate you from everyone in your niche.

Why it matters: Your point of view is your brand. “More tools ≠ more productive” isn’t a tagline, it’s a filter that shapes every sentence. Feed Claude your beliefs and it stops writing safe, agreeable mush and starts writing your angle.

3. Audience — who you’re writing to, in their words

Who you serve and the exact language they use to describe their pain.

Why it matters: Writing “to everyone” produces writing that lands on no one. When Claude knows your reader says “I never know how much I’m making next month,” it writes to a person…not into the void. Use their real phrases, not your polished summary of them.

4. Voice & Tone Rules — the do’s and don’ts, as commands

Direct instructions: “Write short. Use real numbers. Never motivate — demonstrate.”

Why it matters: This is your actual style control panel. The trick is to make them imperative and specific. “Be conversational” tells Claude nothing. “Use second person. Max four sentences per paragraph. Lead with the number.” tells it everything.

5. Vocabulary — words you use, words you ban

Two lists: your signature words, and your hard-banned words.

Why it matters: The banned list defines your voice as much as the used list. Every AI creator reaching for “delve,” “unlock,” “leverage,” “supercharge,” “in today’s fast-paced world” is why AI writing has a smell. Ban them explicitly and half your “sounds like AI” problem disappears in one line.

6. Writing Samples — show, don’t tell

Two or three real excerpts of your best writing.

Why it matters: This is the single highest-leverage section. A list of adjectives (”witty, warm, direct”) is weak. One paragraph of you actually being witty, warm, and direct is worth more than a page of descriptions. Claude pattern-matches to real examples far better than to labels. If you only nail one section, nail this one.

7. Formatting Preferences — how your content is shaped

Structure, length, emoji use, how you handle lists and headers. Why it matters: Voice isn’t only word choice…it’s shape.
If you never use emojis and Claude sprinkles them in, it reads as fake instantly. Lock the shape and the output stops feeling borrowed.

8. Offers & Links — your products and canonical URLs

Your offers, prices, lead magnet, and the exact links you use.

Why it matters: The moment you ask Claude to write a CTA, it either invents a link or asks you for one. Put them here once and every call-to-action comes out correct and on-brand.

9. Hard Rules — the non-negotiables

A short list of never-do lines. “Never fabricate stats. Never use hype. Never write a fake personal story.”

Why it matters: Reserve this for what truly matters. If you mark everything important, nothing is. Three genuine hard rules get followed. Twenty “important” rules get averaged into noise.

That’s the skeleton.

Now let’s put meat on it…starting from wherever you actually are today.

The Build: Three On-Ramps

You’re in one of three places right now.

Find yours and start there.

All three roads meet at the same file.

On-Ramp A: “I don’t have a brand yet — I’ve barely started”

No posts, no positioning, maybe you’ve never seriously thought about building online. Perfect.

You have no bad habits to unlearn.

Don’t try to invent a brand in your head. Extract it. Open Claude and paste this:

Read more

Not a “what do I post next” problem — the opposite. It’s the pile of stuff you already made. Posts, essays, newsletters that did one lap, got their moment, and went quiet. You spent hours on an article, it landed, and now it’s just… sitting there. Gathering dust.

The usual advice is “repurpose it…turn that post into a video.” Sure. Except turning one article into a TikTok-ready video by hand is its own part-time job: script, visuals, captions, editing. So most of us never do it. The archive keeps growing. The dust keeps settling.

Here’s my take on that: the wrong answer is “hustle harder.” The right answer is a system…something you build once that turns a task you’d never do manually into a single command you actually run. As Vinayak puts it in the piece: you don’t build a habit, you build a machine that runs the habit for you. That’s the whole game.

That’s why this guest post stopped me.

Vinayak Ramesh had 87 articles gathering dust. A full-time research job, a Substack, a Medium account he posts to three or four times a week…and a backlog he was never realistically going to turn into video by hand. So he built a pipeline in Claude Code that does it for him: drop in any article, run one terminal command, and 8–15 minutes later you get back a finished vertical video plus a ready-to-paste caption, hashtags, and Pinterest description. His active time per video: about 30 seconds.

Not a tool he bought. A system he built. One command.

Vinayak writes The Sustainable Solopreneur, a weekly newsletter for people building income streams alongside a 9-to-5…real experiments, honest results. He also shares templates, prompt packs, and playbooks on Gumroad.

Before we start…

This post is brought you by, Revid.AI

Revid.ai is an AI-powered video production platform trained on millions of viral videos to generate high-performing scripts and scroll-stopping visuals.

From product demos and explainer videos to social ads

Try for Free ✅

back to the post… 👇

In this post Vinayak Ramesh walks through the whole build…the one-command demo and what actually fires in those 8–15 minutes, why he skipped CapCut and Zapier and built it in Claude Code instead, a real unedited first run (script, images, and post package straight out of the pipeline), and a full step-by-step so you can rebuild it for your own archive, copy-paste prompts and hard-won bug fixes included.

One detail worth reading closely: he runs it mostly for Pinterest, not TikTok…because a pin keeps sending traffic back to the article for months while a Reel is dead in four hours. That reframe alone is worth the read.

If you’ve got a folder full of posts you’re quietly ignoring, this is the system to steal.

Here’s Vinayak. 👇

I have a full-time research job. A Substack newsletter. A Medium account where I post three or four times a week. And somewhere in the last few months I realized I was sitting on 87 articles that had done their job once and then just... stopped.

They’re not bad pieces, some of my best writing lives in that pile. But turning any one of them into a video for YouTube, Instagram, TikTok, or Facebook meant starting from zero every time: new script, new visuals, new captions. I don’t have three extra hours a night for that. Not with a full-time job.

So I stopped trying to create more content and started asking a different question: how do I get more mileage out of the content I already have?

That’s the whole premise of this pipeline. It doesn’t write anything new. It takes an article I already published on Medium or Substack and turns it into a short-form video…same idea, same message, repackaged for a platform where people scroll instead of read. More places for people to find writing I already did the hard work on. More of those people eventually clicking through to the article, subscribing to the newsletter, sticking around.

Right now I’m running it mainly for Pinterest, on purpose. Instagram/TikTok is a firehose…a video is basically dead within 4 hours and you’re back to zero. Pinterest works more like a search engine: a decent pin keeps sending traffic back to an article for months. If I’m putting in zero extra creative energy to repurpose something, I want that repurposed version working for me long after I’ve forgotten I posted it.

Here’s what actually runs when I want a new video. One command:

TERMINAL

npm run create -- --daily

What happens next is it takes 8–15 minutes and needs zero input from me.

Claude Code reads one of my articles. Gemini AI writes a 50–60 second video script with 9–11 scenes. Imagen 4 generates an illustration for every scene. Remotion renders a polished 1080×1920 MP4 with animated captions, Ken Burns motion effects, and scene transitions. A metadata generator writes the complete TikTok caption, hashtags split across caption and first comment, Pinterest title, and description — formatted and ready to copy-paste.

My active time: 30 seconds to check the output.

This article is everything — every piece I built, every mistake that cost me an evening, and exactly how you can put the same thing together for whatever you’re already writing.

Why Is Repurposing So Hard to Do Consistently?

The bottleneck is never creativity. It’s activation energy.

The gap between “I should turn this into a video” and actually sitting down to do it is just large enough that it never happens consistently. Most creators solve this by trying to build better habits — scheduling time, writing scripts in advance, batching on weekends.

The only way to solve an activation energy problem is to remove the activation step entirely.

That’s what this pipeline does. It doesn’t help you make videos faster. It removes you from the video-making process almost completely. The only decision you make is whether the output is good enough to upload.

You don’t build a habit. You build a machine that runs the habit for you.

Why I Built This Instead of Using Existing Tools

Before building, I actually tried to find a shortcut.

The manual process for one video was roughly this: finish an article, open Notion and write a script doc, generate each scene image in ChatGPT one by one (9–11 separate prompts), download the images, drag them into CapCut, set timing, add captions by hand, export, then switch back to ChatGPT to write the TikTok caption and hashtags. Start to finish: 45–90 minutes per video. At one video per day, that is 22–45 hours per month just on repurposing.

I looked at alternatives. CapCut has AI features but needs manual input for every video. Descript is great for editing recorded content but does not take a text article and produce a short-form video without a recording. Make and Zapier workflows would have worked but needed a server, API keys wired together, and ongoing maintenance — three things that break at the worst possible time.

None of them do all seven things: read the article, decide which article to use, write the script, generate images, render the video, write the post package, and log the history. That is the full pipeline. Anything that only does two or three stages still leaves the other stages on your plate.

Claude Code handles the whole thing in one session of natural-language prompting. No server. No wiring. No maintenance beyond what Claude Code manages itself. That was the reason.

What Does One Command Actually Do?

When npm run create -- --daily runs, seven stages fire in sequence:

Two files come out: one .mp4 and one .txt post package. That is everything needed to upload.

What Does It Cost to Run?

Getting your Gemini API key: Go to aistudio.google.com. Paid usage is billed per image and per token.

What Did It Actually Produce on Its First Run?

This isn’t a hypothetical. This is a real video, still live, built from an article of mine called “Why Your KDP Book Isn’t Selling.” The pipeline read the article’s tone and picked the STORY_ARC style on its own, with a PERSONAL_CONFESSION hook — problem, struggle, insight, result, told like a confession instead of a lecture. Here’s exactly what the script generator wrote:

SCRIPT BREAKDOWN — REAL OUTPUT

Article:  Why Your KDP Book Isn’t Selling
Style:    STORY_ARC
Hook:     PERSONAL_CONFESSION
Duration: 62s | 11 scenes

Hook (first 3s): “My books failed.”

All captions:
[1]  “My books failed.”
[2]  “Not bad content. Not ugly covers.”
[3]  “They just sat there, invisible.”
[4]  “The difference wasn’t quality.”
[5]  “It was one decision *before* writing a word.”
[6]  “I wrote what I *wanted* to.”
[7]  “Smart publishers find what the market *needs*.”
[8]  “Your book needs a specific, searchable keyword.”
[9]  “Validate that keyword with real market data.”
[10] “Look for niches with sales *and* low competition.”
[11] “This one shift guarantees your book gets found.”

Imagen 4 generated these directly from that script — no touch-ups, no picking favorites, this is what came out of the pipeline on the first pass:

Scene 1 — “My books failed.”

Scene 3 — “They just sat there, invisible.”

Scene 8 — “...a specific, searchable keyword.”

Actual scene stills, 9:16, straight out of the pipeline — no edits.

Remotion assembled all 11 into the MP4. Then the metadata generator wrote this — also real, also unedited:

METADATA GENERATOR OUTPUT — REAL OUTPUT

TIKTOK / REELS

On-Platform Title:
My KDP Books Flopped: The Invisible Reason Yours Aren’t Selling

Caption:
My KDP books were amazing, but nobody bought them. Turns out, content
quality wasn’t the issue. Discover the secret to making your KDP books
visible and profitable. Follow for more solopreneur insights!
#TheSustainableSolopreneur #stickfigure #solopreneur #newsletter
#tiktoksolopreneur #reels #tiktokcreator #substackgrowth #onlinewriting
#buildinpublic

First Comment (post immediately after):
#KDP #SelfPublishing #AmazonKDP #BookMarketingTips #IndieAuthor
#KindleBooks #AuthorLife #BookLaunch #DigitalProducts #PassiveIncomeBooks

PINTEREST

Title:
KDP Book Not Selling? The #1 Reason It’s Invisible (And What To Do)

Description:
Your KDP book might have great content and a stunning cover, yet it’s
gathering dust. The problem isn’t quality, it’s visibility. Discover why
many self-published books fail to sell and the key difference that
drives sales. Stop letting your hard work go unseen.
#TheSustainableSolopreneur #stickfigure #solopreneur #newsletter

Active time from command to a finished post package: under fifteen minutes, almost all of it spent watching the terminal scroll. My time involved: pressing enter, then hitting upload once the file was ready.

I opened the MP4, copied the Pinterest title and description straight out of the .txt, and uploaded. It’s been sitting on Pinterest since, quietly doing its job.

This exact video, live on Pinterest

---

How Does the Pipeline Get Smarter Over Time?

The first run treats every article as a fresh start. By week three, the pipeline has compounding logic working in three directions.

The article picker learns recency. It tracks every article that has been turned into a video, with the date. After 14 days, a used article becomes eligible again — but with freshAngle: true set, which tells the script generator to find a completely different angle on the same content. One article can become three or four videos over months without repeating itself.

The history tracker learns cadence. shouldPromoteGumroad() returns true every fifth video (totalVideos % 5 === 0). On those runs, the metadata generator includes a product mentioned in the caption automatically— something like “If you want the exact framework, grab the Substack Growth Engine“ with the link swapped in…You never have to remember to promote your products. The pipeline rotates it on a schedule.

The style logic learns variety. The first version cycled the same 10 visual types in a fixed rotation. I expanded it to 20 types with emotion-register-aware selection — no two adjacent scenes share the same type, the first scene always uses a high-emotion hook type, and the last scene always uses a payoff type. Every run looks different from the last one.

Day one it is a production machine. Month three it is an efficient production machine that knows your content catalog.

Inside the Pipeline: How Each Stage Works

How Does the Article Picker Decide What to Use?

The picker follows a strict priority order:

1. Never-used articles — highest priority

2. Rested articles — used more than 14 days ago, eligible again

3. Recently used — last resort fallback only

It also weights by platform: 70% Medium, 30% Substack. Short-form video drives more referral traffic back to Medium, so Medium articles get first pick.

How Does the Script Generator Work?

The cleaned article text goes into Gemini 2.5 Flash with two settings that most people get wrong on the first attempt:

TYPESCRIPT

responseMimeType: “application/json”  // forces structured output, no markdown
wrapping
maxOutputTokens: 8192                  // prevents mid-JSON truncation

Without responseMimeType: “application/json”, Gemini wraps the output in markdown fences. Without maxOutputTokens: 8192, it truncates mid-JSON on longer scripts and throws a parse error. Both mistakes cost debugging time I already paid for.

The model returns a hookStyle and a scenes array. Each scene contains durationSeconds, caption (max 15 words), optional subCaption (max 8 words), imagePrompt, and visualType.

The pipeline chooses from 6 video styles based on the article’s tone and structure:

20 visual types are assigned by an algorithm — not a fixed rotation. First scene: always a high-emotion hook type (SOLO_REACTION, COUNTDOWN_CLOCK, SPOTLIGHT_SOLO, MEME_ROCKBOTTOM). Middle scenes: rotate through emotional registers — no two adjacent scenes share the same type. Last scene: always a payoff type (CELEBRATION, MOUNTAIN_SUMMIT, STAT_HERO).

Every image prompt also gets a perspective layer (one of 8 camera angles) and an atmosphere layer (one of 8 background treatments) added automatically. This means two videos with the same visual type will still look compositionally different.

How Does the Image Generator Avoid Bad Output?

Each scene’s prompt goes to Imagen 4 first, with a fallback if it fails.

Primary model:

TYPESCRIPT

const result = await ai.models.generateImages({
model: “imagen-4.0-generate-001”,
prompt: scene.imagePrompt,
config: { numberOfImages: 1, aspectRatio: “9:16” }
});

Fallback model:

TYPESCRIPT

const result = await ai.models.generateContent({
model: “gemini-3.1-flash-image”,
contents: scene.imagePrompt,
config: { responseModalities: [”IMAGE”, “TEXT”] }
});

There is a 600ms delay between calls to avoid rate limiting. Images are cached to disk — if the pipeline crashes mid-run, already-generated images are skipped on restart.

Three rules that took the most trial and error to figure out:

• Never use hex codes in image prompts. #7DC242 will appear as literal floating green text. Use plain English instead: “vivid neon lime green.”

• Never describe text or letters in scene compositions. Even “a sign with an X” will cause the model to render actual characters.

• End every image prompt with: “CRITICAL: absolutely zero text, numbers, hex codes, or letters anywhere in this image.”

How Does Remotion Render the Video?

Remotion renders React components into video frames. Output: 1080×1920, 30fps, H.264.

The one thing that breaks every first-timer’s setup: Remotion runs inside a browser sandbox and cannot load file:// paths. Every image needs to be converted to a base64 data URI before being passed to the React component. This happens in scenesToDataURIs() before rendering starts. Skip this step and your video renders as black frames.

Each scene in Scene.tsx gets a full-frame image, a Ken Burns effect (8 preset zoom/pan variations cycling per scene), bottom and top vignettes, a watermark in the top-left corner at 40% opacity, and a 9-frame fade in/out at each scene boundary.

Caption animation in Caption.tsx rotates through 3 styles: word-by-word (synced to frame timing), highlight (current word glows in your accent color), and all-at-once (spring animation slide-in for the full line). A lime green accent bar animates on entry. SubCaption sits below in your brand color.

What Does the Metadata Generator Produce?

Once the video renders, one final Gemini call writes the complete post package:

• videoTitle — TikTok/Reels on-platform title

• captionBody — 2–3 sentence hook caption

• 20 hashtags split across two groups: first 10 in the caption, next 10 in the first comment (TikTok best practice for reach)

• firstComment — the second 10 hashtags, paste immediately after uploading

• pinterestTitle — SEO-optimized curiosity-gap hook

• pinterestDescription — 2–4 sentence tease with hashtags

Everything formats into a .txt file with clearly labeled sections. No editing required before posting.

How Do You Build This From Scratch?

Step 1 — Create the Project

Inside your content vault or any folder:

TERMINAL

mkdir _VideoMaker && cd _VideoMaker
npm init -y

Where Do Your Articles Actually Come From?

I use an Obsidian vault, so article-loader.ts reads .md files out of a folder and article-picker.ts chooses between them automatically. That’s my setup — it doesn’t have to be yours.

The pipeline doesn’t care where your content lives. All article-loader.ts needs is a string of plain text. Three ways people actually do this:

1. You already have a vault or folder of articles (Obsidian, a content/ folder, whatever). Build article-loader.ts and article-picker.ts exactly as described in Step 6 — point them at that folder, and the picker rotates through everything in it automatically.

2. Your content lives in Notion. Export the page as Markdown (Notion: ••• → Export → Markdown & CSV), drop the exported .md file into a folder, and point article-loader.ts at that folder instead. Nothing downstream changes — Gemini doesn’t know or care that the text came from Notion.

3. You just want to paste one article and go. Skip the picker logic entirely. Copy your published post straight from Substack or Medium, paste it into a single .txt file, and tell Claude Code to build a stripped-down article-loader.ts that just reads that one file — no priority queue, no rotation, no --daily flag. Run it with npm run create -- --article “your-file-name” whenever you have something new to repurpose.

If you’re only repurposing a handful of pieces a month, option 3 is genuinely the easier build. The article picker’s 14-day recency logic only starts paying for itself once you have a real backlog — mine only became worth automating after I hit 87 unused articles.

Step 2 — Install Dependencies

TERMINAL

npm install \
remotion@4.0.481 \
@remotion/bundler@4.0.481 \
@remotion/renderer@4.0.481 \ 
@remotion/cli@4.0.481 \ @google/genai \
tsx gray-matter dotenv \
react react-dom \
@types/react @types/react-dom \
typescript

⚠️ All four Remotion packages must share the same version number. Do not install @remotion/core separately — it has a different versioning scheme and will break your build.

Step 3 — Configure TypeScript

Create tsconfig.json:

TSCONFIG.JSON

{
“compilerOptions”: {
  “target”: “ES2020”,
  “module”: “CommonJS”,
   “moduleResolution”: “node”,
   “jsx”: “react”,
   “esModuleInterop”: true,
   “strict”: true,
   “outDir”: “dist”
 },
“include”: [”src”]
}

Step 4 — Add Your API Key

Create .env in the project root:

.ENV

GEMINI_API_KEY=your_gemini_api_key_here

Add the run script to package.json:

PACKAGE.JSON

“scripts”: {
”create”: “tsx src/cli.ts”
}

Step 5 — Define Your Brand Guidelines Before Writing Any Code

This is the step most tutorials skip and it is the most important one.

The pipeline encodes your brand into every image prompt and every rendered frame. If you skip this, all your videos will look generic. Before you touch the code, answer four questions:

4. What is your visual style? The illustration style for every scene image. Minimalist stick figures, flat vector illustration, cartoon characters, geometric shapes, photorealistic backgrounds.

5. What are your brand colors? You need three values — background, accent (captions, highlights, bars), and text. Always describe colors in plain English inside prompts. Never use hex codes — image models render them as literal on-screen text.

6. What is your watermark? Your handle or brand name, shown in the top-left corner of every scene at 40% opacity.

7. What is your target video duration? Default is 50–60 seconds with 9–11 scenes. Adjust in the script generator’s system prompt for shorter or longer content.

Once you have your answers, paste this into Claude Code:

COPY-PASTE PROMPT

I’m building a short-form video pipeline. Before we write any code,
I need to define my brand guidelines so Claude can encode them into
every image prompt and rendered frame.

Visual style: [describe your illustration style]
Background: [describe in plain English — e.g. “deep solid black”]
Accent color: [describe in plain English — e.g. “vivid neon lime green”]
Text color: [e.g. “bright white”]
Watermark: [@YourHandle]
Target duration: [e.g. “50–60 seconds, 9–11 scenes”]
Create a brand guidelines object I can reference throughout the
pipeline build. Write the image prompt style description in plain
English — no hex codes inside prompts.

Here is what mine looked like:

MY FILLED EXAMPLE

My visual style: Flork-style minimalist stick figures — simple,
expressive, slightly chaotic energy. Every scene should feel like
a comic panel.

Background: Deep solid black
Accent color: Vivid neon lime green
Text: Bright white
Watermark: @TheSustainableSolopreneur
Duration: 50–60 seconds, 9–11 scenes
One rule: never use hex codes in image prompts. Describe every
color in plain English. Always end every image prompt with:
CRITICAL: absolutely zero text, numbers, hex codes, or letters
anywhere in the image.

Claude already knew what Flork-style stick figures looked like from its training. From that single briefing, every image prompt it generated for the rest of the session described scenes in my exact visual language without me repeating the brief each time. Brief once. The pipeline carries it forward.

What If You Don’t Want Stick Figures?

Everything in this article uses my brand — Flork-style stick figures, black background, lime green accents because that’s what fits my brand. None of that is hardcoded into the pipeline. It all comes from the one paragraph you wrote above.

Swap that paragraph and every image prompt the pipeline generates changes with it, because the script and image generators both inherit whatever visual style description you gave Claude Code — they don’t have a fixed style baked into the code. A few descriptions that work well if stick figures aren’t your brand:

• Flat vector illustration: “Flat 2D vector illustration, bold geometric shapes, limited 4-color palette, no gradients, editorial explainer style.”

• 3D isometric: “Isometric 3D render, soft studio lighting, rounded plastic-toy aesthetic, single accent color, clean shadows.”

• Paper-cutout collage: “Layered paper-cutout collage style, visible paper texture and drop shadows, muted craft-paper palette.”

• Photoreal B-roll style: “Photorealistic stock-photo style, shallow depth of field, natural lighting, no illustrated elements at all.”

• Anime / manga: “Clean manga-style line art, screentone shading, expressive faces, monochrome with one spot color.”

Whichever you pick, keep the same three rules from above: describe every color in plain English, never ask for on-screen text or letters, and end every prompt with the zero-text instruction. Those rules aren’t stick-figure-specific — they’re about what image models get wrong regardless of style.

Step 6 — Build with Claude Code

Build each module in this order:

BUILD ORDER

1.  types.ts
2.  article-loader.ts
3.  script-generator.ts
4.  image-generator.ts
5.  renderer.ts
6.  article-picker.ts
7.  history-tracker.ts
8.  metadata-generator.ts
9.  cli.ts
10. remotion/components/Scene.tsx
11. remotion/components/Caption.tsx

Describe what you want in plain English for each module. Claude Code handles the TypeScript types, API call structure, error handling, and JSON parsing. You correct the output, iterate, and move to the next module. The entire build took me one session of about 2–3 hours, including all the debugging.

Step 7 — Run It End to End With Your Own Article

Everything above gets you a working pipeline. Here’s what to actually do with it, start to finish, the first time.

1. Drop one of your own articles into your content folder. Whatever you decided above — a .md file from your vault, a Notion export, or a .txt file pasted straight from a live post. Use something you’ve already published; you want to judge the output against writing you already know well.

2. Run the script-only pass first. npm run create -- --article “your-article-name” --script-only. This skips images and rendering entirely and just prints the script JSON — hook style, video style, and every scene’s caption. Read it before you spend a cent on image generation. If the captions feel off, that’s a script-generator prompt problem, not an image problem — fix it here.

3. Run the full command once the script looks right. npm run create -- --article “your-article-name”. Watch the terminal: the article gets picked up, the script generates, then a run of image-generation calls (roughly one every 600ms–2s depending on the API), then the Remotion render logs, then the metadata generation, then a final line confirming the save location.

4. Open the output folder. You’ll have two files: the .mp4 and the -post-package.txt. Watch the video first, full screen, sound off — most people watch short-form muted, so if the captions alone don’t tell the story, fix that before you post.

5. Copy the post package and upload. The .txt file already has your title, caption, hashtags, and Pinterest title/description written out. Paste, don’t rewrite — that’s the entire point of the pipeline.

6. Log what worked. Nothing in the code does this part for you. After each upload, jot one line — which hook style landed, which visual type felt flat — in a plain notes file. That’s the manual half of the feedback loop from the brand-brief step, and it’s the difference between a pipeline that stays static and one that gets sharper every week.

The Full Folder Structure

FOLDER STRUCTURE

_VideoMaker/
├── package.json
├── .env
├── src/
│   ├── cli.ts                     ← entry point, orchestrates all stages
│   ├── types.ts                   ← shared TypeScript interfaces
│   ├── remotion/
│   │   ├── index.ts               ← registers the Remotion composition
│   │   └── components/
│   │       ├── Scene.tsx          ← image + Ken Burns + vignette + watermark
│   │       └── Caption.tsx        ← animated caption component
│   └── pipeline/
│       ├── article-loader.ts      ← reads .md files, strips CTAs
│       ├── script-generator.ts    ← calls Gemini, returns structured script
│       ├── image-generator.ts     ← calls Imagen 4, caches to disk
│       ├── renderer.ts            ← Remotion bundler + renderMedia
│       ├── article-picker.ts      ← priority-weighted article selection
│       ├── history-tracker.ts     ← logs videos, tracks cadence
│       └── metadata-generator.ts  ← TikTok + Pinterest post package
├── temp/                          ← working images per job (auto-cleared)
├── output/
│   └── 2026-06-25/
│       ├── article-name.mp4
│       └── article-name-post-package.txt
└── logs/
    └── video-history.json

Useful Copy-Paste Prompts to Start

Four prompts run the whole system once it’s built. Paste them as-is.

TERMINAL

# Daily auto-pick
npm run create -- --daily
 
# Specific article by name (swap in your own title/filename)
npm run create -- --article "Your Article Title Here"
 
# Force a style
npm run create -- --daily --style MYTH_BUSTER
 
# Preview just the script structure (fast, no images, no render)
npm run create -- --daily --script-only
 
# Generate images only, skip render
npm run create -- --daily --no-render
 
# List all available articles
npm run create -- --list

What Breaks and How to Fix It

The full detail on each problem and what caused it:

Problem 1 — Hex codes appearing as text in images. Imagen 4 rendered #7DC242 as literal floating green text. Fix: plain English colors only, plus the CRITICAL no-text instruction at the end of every prompt.

Problem 2 — JSON truncation mid-output. Error: Expected ‘,’ or ‘]’ after array element. Fix: responseMimeType: “application/json” to force structured output, AND maxOutputTokens: 8192. The default limit was too low for 11-scene scripts.

Problem 3 — Black frames. Images were passed as file:/// paths. Fix: convert everything to base64 data URIs in scenesToDataURIs() before calling renderMedia().

Problem 4 — Process killed after rendering. Remotion’s onProgress fires on every frame — 1,800 calls for a 60-second video. Writing to stdout 1,800 times overflowed the buffer, killing the process before the post package saved. Fix: throttle to ≥5% increments, add logLevel: “error”.

Problem 5 — Wrong model names. gemini-2.0-flash was deprecated. Fix: use gemini-2.5-flash for text and imagen-4.0-generate-001 via generateImages() for images.

Problem 6 — Imagen 4 safety filter error. Error: Only block_low_and_above is supported. Fix: remove the safetyFilterLevel parameter entirely. Don’t pass it at all.

Problem 7 — Article loader stripping too much content. The CTA stripper matched on the keyword “subscribe” and deleted entire paragraphs that used the word in passing. Fix: target the specific CTA block structure — the three-paragraph block at the end of each post — rather than individual keywords anywhere in the article.

Problem 8 — Every video looking identical. The first version cycled the same 10 visual types in a fixed rotation. Fix: expand to 20 types, add emotion-register-aware selection (no adjacent repeats, hook-specific first scene, payoff-specific last scene), and add perspective and atmosphere layers to every image prompt.

What Mistakes Should You Avoid?

The four that quietly break the system:

1. Skipping the brand brief. Defining your visual style, colors, and watermark before writing any code is what makes every video look like yours. Skip it and you get generic output that could belong to anyone. The brief takes ten minutes. Skip it and you waste three hours on output you do not want to post.

2. Using hex codes in image prompts. This one will burn you on the first run. #7DC242 becomes floating on-screen text. There is no workaround — plain English is the only option. Write it into every single image prompt, not just the system instructions.

3. Running Remotion without base64 conversion. It renders black frames silently. No error. No warning. Just a black video at the end of a 12-minute run. Always call scenesToDataURIs() before renderMedia().

4. Skipping the per-run notes on what worked. The history tracker tells the pipeline what has been done. Your own notes — which outputs were strong, which were mediocre, what you changed — are the feedback loop that makes the brand brief sharper over time. Skip them and the pipeline stays exactly as smart as day one.

Key Takeaways

• The bottleneck is activation energy, not creativity. You don’t need better habits. You need a machine that removes the activation step entirely.

• Claude Code handles the full build in one session. Describe each component in plain English, iterate on the output, move to the next module. Two to three hours, start to finish.

• The article picker has compounding logic. Unused articles go first. After 14 days, used articles become eligible again with freshAngle: true — a different angle on the same content. One article can become multiple videos over time.

• Never use hex codes in image prompts. Image models render them as literal on-screen text. Plain English only. Add the CRITICAL no-text instruction to every prompt.

• `responseMimeType: “application/json”` and `maxOutputTokens: 8192` are non-negotiable. The defaults fail on longer scripts. Set them before your first run.

• Convert all images to base64 before Remotion. Local file paths don’t work in its browser sandbox. Black frames are the symptom.

• Define your brand before writing a single line of code. Brief Claude Code once. It carries the brand context through the entire session without you repeating it.

The One Thing to Do This Week

Don’t try to build the full pipeline today.

Start with Step 5 — define your brand guidelines. Answer the four questions. Write the brief. Open Claude Code and paste it.

Then prompt the script generator as the first module:

COPY-PASTE PROMPT

Build script-generator.ts. It takes a string of article content
and calls Gemini 2.5 Flash with responseMimeType: "application/json"
and maxOutputTokens: 8192 to return a structured video script JSON.
 
Output format:
{
  hookStyle: string,
  videoStyle: "PUNCHY_CUTS" | "STORY_ARC" | "MYTH_BUSTER" |
              "RAPID_FIRE" | "DEEP_REVEAL" | "LIST_FORMAT",
  scenes: [{
    durationSeconds: number,
    caption: string,
    subCaption?: string,
    imagePrompt: string,
    visualType: string
  }]
}
Apply these brand guidelines to every imagePrompt:
[paste your brand brief here]
 
Set up the project with package.json, tsconfig.json, and .env
for GEMINI_API_KEY.

Paste one of your articles in. Run it. Read the script output.

If the scripts look right, add the image generator next. Then Remotion. Build stage by stage.

You already have the content. You already have Claude Code. The only thing missing is the pipeline.

If you build this, I want to see what it produces for your content — comment here or DM me on Substack.

One last thing from me

The part of Vinayak’s build I keep coming back to isn’t the pipeline. It’s the reframe:

the bottleneck was never creativity…it was activation energy.

And the fix wasn’t more discipline. It was a machine that deleted the step entirely.

That’s the whole thing I hammer on here. Systems, not hustle. Build it once, let it run.

You don’t need 87 articles to start. You need one post you’ve already published and a free afternoon. Do exactly what Vinayak says: don’t build the whole pipeline today…just write the brand brief and get the script generator working.
That’s the first domino. Everything downstream falls from there.

If this was useful, go subscribe to Vinayak at The Sustainable Solopreneur…real experiments, honest results, the kind of tested building this newsletter runs on.

And if you want more systems built exactly like this…lean, one-command, doing the work while you sleep…that’s the entire point of The Lean AI-Powered Business Playbook for Creators. It’s how you go from “I should automate that” to actually running it.

One more thing. If you’ve been getting value from these breakdowns and want the full playbooks, the build-alongs, and everything I don’t publish for free, here’s why it’s worth upgrading to premium. Come build with me on the inside.

— Sharyph

Weeks? Months? Maybe years?

And how much have you earned from it?

If there’s a gap between those two answers…you’re not alone.

Here’s what I see constantly in this space:

Creators who are genuinely talented. Writing threads. Publishing newsletters. Building audiences. Putting in real work every single day.

But the revenue? Flat. Inconsistent. Sometimes nonexistent.

And they can’t figure out why.

Here’s the truth nobody wants to say out loud:

I’ve been there. When I started, I had ideas. But no direction.

No clear offer. No system. Just noise.

So let’s fix that today.

Because when I looked at how creators like Justin Welsh crossed $5M in revenue, how Lenny Rachitsky built a $2M+/year paid newsletter, and how Easlo turned Notion templates into $500K+ at 20 years old…

One pattern showed up every single time.

Not luck. Not going viral. Not some secret platform hack.

A ladder.

A specific, repeatable product ladder that turns your knowledge into income…rung by rung.

And the best part?

AI now compresses every single step of that ladder.

What used to take years can now take months. What used to require a team can now run with one person and a $100/month AI stack.

That’s what this issue is about.

Here’s exactly what you’ll get:

✅ The hard truth about why most creators never earn consistently (and the data to back it up)

✅ The 4-rung product ladder that every 6 and 7-figure creator has climbed

✅ An 8-step repeatable framework you can start executing this week

✅ The AI execution layer…which tools to use, at which step, and how

✅ The reality filter…so you go in with the right expectations

This is not theory.

This is the playbook.

Let’s get into it. 👇

The Truth About Creator Revenue (Read This Before You Do Anything Else)

Before we get into the playbook…

I need to tell you something the creator economy doesn’t want you to hear.

Because if you skip this part, nothing else in this issue will land the way it should.

The Numbers Behind the Dream

You see the screenshots.

The “$10K month.” The “I quit my job.” The “passive income while I travel.”

And they’re real. Some of them, anyway.

But here’s what nobody shows you alongside those screenshots:

The results?

❌ 44% of products on the platform earned exactly $0.

❌ The median product earned just 28 sales at a $13 price point…roughly $364 lifetime.

❌ The top 1% of sellers control 99.5% of all revenue on the platform.

Read that again.

The top 1% take nearly ALL of it.

That’s not a platform problem. That’s a positioning problem.

And the creators sitting in that top 1%? They didn’t get there by just posting more. They got there by doing something structurally different from everyone else.

The Wrong Order is Killing You

Here’s the mistake I see constantly.

Creator gets inspired → Creates a product → Tries to find an audience to sell it to.

That’s backwards.

Every single 6 and 7-figure creator in this space did it the other way around.

They built an audience first.
Gave away free value until people trusted them.
Then monetized that trust with a product that solved a real, documented problem.

👉 Justin Welsh posted on LinkedIn every single day…for years…before his first course.

👉 Lenny Rachitsky wrote weekly on Substack for 9 full months before he charged a single dollar.

👉 Easlo shared free Notion templates on Twitter and Product Hunt before he ever put a price tag on anything.

The product didn’t create the audience. The audience validated the product.

That’s the order. And if you get it wrong, it doesn’t matter how good your product is.

The One Pattern That Actually Explains 6 and 7-Figure Creators

I spent a lot of time studying the creators who crossed six and seven figures.

Justin Welsh ($5M+ one-person business).
Ali Abdaal ($4.5M–$5M/year).
Lenny Rachitsky ($2M+/year from his newsletter alone).
Dan Koe (~$2.5M in 2023).
Easlo ($500K+ from Notion templates).

They are completely different people. Different niches. Different platforms. Different product types.

But strip away all of that, and the same pattern shows up every single time:

It doesn’t matter if you sell templates, courses, a community, or a paid newsletter.

What matters is that you built a real audience on ONE platform first…then converted that audience into income using a structured product ladder.

Justine Welsh built on LinkedIn.
Easlo and Nicolas Cole built on X.
Ali Abdaal built on YouTube.
Lenny built on Substack.

One platform. Consistent free value. A specific niche. Then a product.

That’s it.

That’s the whole game.

But Here’s the Catch

Knowing the pattern isn’t enough.

Because the real question isn’t what they did.

It’s how they did it…step by step…in the right order.

That’s what the rest of this playbook is going to show you.

And we’re going to layer in exactly how AI compresses every single step of the process.

Because the creators who figure this out right now…in 2025 and 2026…are going to climb this ladder in months, not years.

Let’s get into it. 👇

The Product Ladder: The Real Business Model Every 6 and 7-Figure Creator Uses

Here’s a question most creators never ask themselves:

“What’s the path from someone discovering me today… to them spending $500 with me in 6 months?”

If you can’t answer that question clearly, you don’t have a business.

You have a content hobby.

The creators who crossed six and seven figures didn’t get there by selling one thing to everyone.

They got there by building a ladder…a structured sequence of products that take someone from free value all the way up to premium.

And here’s the insight that changes everything:

Every rung of the ladder serves a different purpose.

The bottom rungs build trust. The middle rungs build revenue. The top rungs build real leverage.

Let me show you exactly how this works.

🪜 Rung 1 — The Entry Point (Months 0–6)

Free Lead Magnet + Low-Ticket Digital Product

This is where every creator should start.

Not with a course. Not with a community. Not with a coaching offer.

With a free lead magnet and a cheap, narrow digital product…priced between $20 and $150.

Why?

Because at this stage, nobody knows you yet.

You haven’t earned the right to ask for $500.
But you can earn the right to ask for $27 or $47…if what you’re offering solves a real, specific problem.

Here’s the proof:

Easlo…earning $500K+ from Notion templates…started by sharing free minimalist templates on Twitter and Product Hunt. He switched from $0 to paid by literally changing the price on Gumroad. His first paid template brought in over $3,000 in the first month.

He didn’t start with a $297 course.

He started with a $0 template. Then a $19 template. Then a $130 template. Then a $299 bundle.

Rung by rung.

What belongs on Rung 1:

✅ A free lead magnet (a guide, checklist, mini-playbook, or template)

✅ A low-ticket digital product ($20–150): a Notion template, ebook, prompt pack, mini-guide, or quick-start toolkit

✅ Published on a simple platform: Gumroad is perfect here (no upfront cost, built-in checkout, handles global payments)

The benchmark to move up: Your first 10 sales. Or 100+ email signups from your lead magnet. Either signals enough trust to build the next rung.

💡 I’ve used Gumroad for most of my product launches — 700+ sales, zero ads. It works because it removes every barrier between you and your first dollar.

🪜 Rung 2 — The Leverage Layer (Months 6–18):

Self-Paced Course ($150–500)

This is where things start to compound.

Once you have proof…once real people have paid you for something and told you it helped…you build the bigger thing.

A self-paced course. A comprehensive guide. A step-by-step system.

This is the highest-leverage product format once you have an audience. Near-zero marginal cost. 80–95% margins. And it keeps selling while you sleep.

Here’s the proof:

Justin Welsh built his entire $5M+ business on self-paced courses priced at around $150 each. Two courses. Simple delivery. Massive distribution.
His LinkedIn OS and Content OS courses alone generated millions…at a margin of 92%.

Dan Koe scaled to ~$2.5M in 2023 with “The 2 Hour Writer” and “Digital Economics”…courses in the $499–$999 range.

Neither of them launched these courses with zero audience.

They built trust first. Then they monetized it.

What belongs on Rung 2:

✅ A self-paced course teaching one complete system or transformation

✅ Priced between $150–500 (don’t underprice — pricing below $30 dramatically underperforms in the data)

✅ Pre-sold before it’s built — this is non-negotiable (more on this in Section 3)

✅ Delivered via a simple platform: Gumroad, Thinkific, Stan Store, or Kajabi

The benchmark to move up: Consistent monthly sales + a growing list of people asking for more depth, community, or direct access to you.

💡 My AI Digital Product Builder course ($67) walks through this exact process from validating your idea to building and selling your first course with AI doing most of the heavy lifting.

🪜 Rung 3 — The Recurring Engine (Months 12–24+):

Membership + High-Ticket Cohort

This is where six figures becomes the floor, not the ceiling.

Because here’s the math nobody talks about:

Six-figure creators on Kajabi average just 309 paying customers.

Not 10,000 followers. Not 50,000 email subscribers.

Just 309 people who paid for something.

Read that again.

At $39/month, 309 people is $12,051/month. That’s $144,612/year.

From 309 customers.

That’s the power of recurring revenue. And that’s why the third rung — a community or membership — is where everything shifts from income to a business.

Here’s the proof:

Lenny Rachitsky started with a free newsletter and a small paid tier. Nine months of weekly writing before he charged a dollar. But once he built the recurring layer, he scaled to $2M+/year from paid subscriptions alone.

Codie Sanchez built the Contrarian Community mastermind at $10,000/year. Dickie Bush and Nicolas Cole’s Premium Ghostwriting Academy crossed $3M in revenue in its first year, built entirely on top of the audience and trust they’d built at Rung 1 and 2.

What belongs on Rung 3:

✅ A membership or community ($10–50/month on Skool, Circle, or Substack) — recurring revenue that compounds every month

✅ A cohort-based course or mastermind ($750–$10,000) — live, high-touch, and your highest-ticket offer

✅ These require real audience and existing proof from Rungs 1 and 2 — don’t try to launch here from zero

The benchmark to keep scaling: Consistent recurring revenue + customers asking for even more direct access, which signals your next high-ticket tier.

💡 This is exactly what my Premium Membership does — giving subscribers 45% off all courses, access to premium Claude Skills, AI agents, and a direct community. Priced at $39/month or $150/year.

🪜 Rung 4 — The Flywheel:

Your Email Newsletter as the Owned Bridge

This isn’t a product rung.

It’s the glue that holds every rung together.

Every single creator in this study…Lenny, Welsh, Sahil Bloom, Easlo, Nat Eliason, Codie Sanchez…used email as their conversion engine.

Not social media.

Not DMs.

Email.

Social platforms distribute your ideas. Email converts them into revenue.

Here’s how the flywheel works:

→ You post free value on ONE platform (X, LinkedIn, YouTube, or Substack)

→ That content drives people to your lead magnet

→ Lead magnet grows your email list

→ Email list converts into Rung 1 buyers

→ Rung 1 buyers become Rung 2 buyers

→ Rung 2 buyers become Rung 3 members and community

→ The community generates testimonials and word-of-mouth that feed back into your social content

And the loop keeps compounding.

This is not a launch strategy.

This is a living system that earns every single day — with or without you posting.

Tools for Rung 4:

✅ Kit (formerly ConvertKit) — best for creators selling products (automation, sequences, tags)

✅ Beehiiv — great for newsletter-first creators focused on growth and ad revenue

✅ Substack — best for discovery network effects (but takes a 10% fee on subscriptions)

The Full Picture: What the Ladder Actually Looks Like

The ladder only works in one direction.

You can’t skip rungs.

Trying to launch a $5,000 mastermind with no audience is like trying to climb a ladder starting from the third rung — there’s nothing under you.

But when you build it rung by rung?

Every step you take upward is supported by everything below it.

Trust compounds. Revenue compounds. The product ecosystem compounds.

And with AI compressing every step of the building process — you can climb this ladder faster than any creator who came before you.

That’s what we’re covering next. 👇

Share

The 8-Step Repeatable Framework: From Zero to Your First $10K

You now know what the ladder looks like.

Now let’s talk about how to actually climb it.

Because knowing the destination is not the same as knowing the route.

This 8-step framework is pulled directly from studying the creators who built to six and seven figures. Strip away the different niches, the different platforms, the different personalities…

And this exact sequence shows up every time.

Let’s walk through it step by step.

Step 1 — Choose a Niche (Specific Outcome × Specific Person)

This is where most creators make their first…and most expensive…mistake.

They go too broad.

❌ “I help entrepreneurs.” “I write about productivity.” “I cover AI tools.”

That’s not a niche. That’s a category.

A niche is a specific outcome for a specific person.

Look at how the best in this space did it:

• Tiago Forte started with: “I help people organize their Evernote without tags.” Not “productivity.” Not “note-taking.” Organize Evernote without tags.

• Justin Welsh started with a hyper-specific audience…burned-out healthcare sales executives…before expanding.

• Lenny Rachitsky deliberately kept his focus on product management for consumer tech companies before eventually broadening.

They started painfully specific.

Then they expanded.

Not the other way around.

Your niche test: Fill in this sentence right now.

“I help [specific person] who struggles with [specific pain] to achieve [specific outcome].”

If you can’t complete that sentence without using words like “entrepreneurs,” “professionals,” or “people who want to grow”…you need to go narrower.

Use Claude to run the Skill Monetization Matrix exercise — it surfaces your monetizable skills, maps them to audience pain points, and helps you find the intersection where your expertise meets real demand.

The Sahil Bloom “10-year test”: Before you commit to a niche, ask yourself…can I still see myself talking about this in 10 years?
If the answer is no, keep narrowing until you find something that genuinely energizes you. Excitement is more sustainable than discipline alone.

Step 2 — Validate Before Building

I’ve built products that felt perfect in theory.

Complete with branding, a landing page, even a full email sequence.

Then I launched…

And sat in silence, refreshing Stripe.

Zero sales. Zero feedback. All that effort wasted.

It was painful…not because I failed…but because I realized I had skipped the one step that would have saved me everything: validation.

Here’s the truth about validation that most creators get wrong:

❌ Getting likes on your idea post is NOT validation.

❌ Someone saying “I’d totally buy this!” is NOT validation.

❌ Positive survey responses are NOT validation.

That’s interest. Not intent.

Their money. Their email. Their time. A DM saying “send me the link now.”

The three fastest validation methods:

Method 1 — The Pre-Order Tweet (Nat Eliason’s approach)
Tweet your course idea with a simple landing page and a price. Nat Eliason did this for his AI app-building course and collected $200,000 in pre-orders in a single week. The tweet was the test. The demand was already there.

Method 2 — The $50 PayPal Pot (Dickie Bush’s approach)
Dickie Bush launched Ship 30 for 30 as a $50 accountability deposit in a Slack channel. Six days after the launch tweet, 50 people had joined. That $50 was real commitment — not a like, not a comment. Real money, real signal.

Method 3 — The “3 Buyers” Rule (Lenny Rachitsky’s approach)
Before building anything, find 3 people who will say: “Yes, I will pay for this. Send me the link.” Not “that sounds cool.” Not “maybe.” Three firm commitments. If you can’t find 3, the offer needs work.

The key question to ask after any validation test:

Did I get commitment — or just compliments?

Compliments are cheap. Commitment is gold.

💡 Want the full validation system?
My AI Digital Product Builder course ($67) walks through the exact process I used to validate Twitter Success Mastery before writing a single word of it…and how to use Claude and Reddit data to find real demand signals in under an hour.

Step 3 — Build Your Audience on ONE Platform

Choose one. Not two. Not three.

One.

Every creator who scaled to six figures picked ONE primary platform and went all-in on it before expanding anywhere else.

• Welsh chose LinkedIn. Daily posts. For years.

• Easlo chose X. Free templates shared consistently.

• Ali Abdaal chose YouTube. First 50 videos were “for him,” not the algorithm.

• Lenny chose Substack. One post per week, every single week.

The platform you choose should match your medium:

The benchmark before you monetize: ~1,000 engaged followers or 500 email subscribers. That’s not a magic number…it’s a trust signal. It means enough people care about what you say that some of them will pay for what you build.

And remember: you’re not posting for the algorithm.

You’re posting to document your journey, build trust, and create a body of evidence that you know what you’re talking about.

Ali Abdaal called his first 50 videos practice. Not content. Practice.

Give yourself that same permission.

Step 4 — Create the MVP (Minimum Viable Product)

Here’s where most creators over-engineer.

They want it perfect before they launch it.

Wrong.

The MVP rule is simple: solve ONE problem completely. Not five problems partially. One problem, completely.

Look at how the best kept it tight:

• Justin Welsh’s courses are 2–3 hours long and priced around $150. Not 40-hour deep dives. Focused, specific, complete.

• Thomas Frank’s “Ultimate Brain” Notion template had one job…replace your fragmented productivity system with one organized workspace. One outcome. $129. Generated $760K in a single year.

• My AI Digital Product Builder course was built around one promise: discover your monetizable skill, Validate demand, Build your product, and Launch it.

The MVP test: Can you describe what your product does in one sentence? If it takes a paragraph to explain what your product covers, it’s too broad.

The build-with-AI shortcut:

You don’t need weeks to build your first product. Using Claude, you can:

→ Draft a 30-page ebook or guide in a weekend

→ Build a Notion template in an evening

→ Create a prompt pack or toolkit in a few hours

The goal isn’t a masterpiece.
The goal is a useful, specific, complete solution to one real problem.

Step 5 — Price With Confidence

This one is uncomfortable for most creators.

So let me give you the data instead of opinions.

Gumroad’s analysis of 146,271 products showed that pricing a product at $30+ correlates with roughly 9x more revenue per product than pricing it under $10.

9x. Not 9%. Nine times.

The pricing principles that work:

✅ Price for the transformation, not the hours. A $47 guide that saves someone 3 months of trial and error is worth $47. A $150 course that helps someone land their first client is worth $150. The value is the outcome — not the page count.

✅ Feel a little uncomfortable with the price. Lenny Rachitsky’s rule: if the price doesn’t make you slightly nervous, it’s probably too low.

✅ Use bundles to raise average order value. Thomas Frank’s bundle outsold his individual products 3 to 1. When you give buyers a clear upgrade path, many of them take it.

✅ The $30 floor. Never price a digital product below $20. The data is clear. Price above $30 and your revenue per product jumps dramatically.

Step 6 — Launch Like an Event

Here’s the difference between posting a product link and actually launching.

A post is: “Hey, I just made this thing. Here’s the link.”

The launch sequence that works:

Week before: Tease the problem your product solves. Share the story behind why you built it. Create anticipation without revealing the product.

Launch day: Announce the product with your best hook, a clear transformation statement, social proof if you have it, and a time-limited early-bird price or bonus. Lenny used a 48-hour discount window. That deadline matters more than you think.

Days 2–5: Share a piece of the product content as free value. Address objections. Share a customer win or testimonial. Each email and post gives people another reason to act.

The Educational Email Course (EEC) funnel — Ship 30’s approach:
Build a short email sequence that teaches something genuinely useful related to your product. At the end of each email, a soft CTA to the product. This sequence runs forever… every new subscriber enters it automatically, whether you launched two years ago or yesterday. It’s the difference between a launch and a permanent sales engine.

💡 I covered how to build this kind of automated email funnel in detail in the AI Digtial Product Builder The system does the selling while you keep creating.

Step 7 — Get Your First 10 → 100 → 1,000 Customers

This is where most creators get impatient.

They launch. They get 3 sales. They declare it a failure and pivot.

Don’t do that.

The path from 0 to 1,000 customers is not a straight line. It’s three distinct phases with three distinct strategies.

First 10 customers…your existing network:
Don’t start with ads. Don’t start with cold outreach.

Start with the people who already know you. Reply to comments. Send DMs to engaged followers. Post in communities where your ideal buyer hangs out. Your first 10 sales come from people who already trust you enough to take a chance.

First 100 customers…consistent content + lead magnet + email nurture:
This is where your flywheel kicks in.

Consistent posts → lead magnet downloads → email subscribers → nurture sequence → product sales.

Lenny’s first two guest posts drove 50% of his first-year subscribers.

Codie Sanchez appeared on ~10 podcasts in two weeks and drove hundreds of new subscribers.

Go where your audience already is…and bring them back to your list.

First 1,000 customers…referrals, affiliates, and platform discovery:
Justin Welsh built an affiliate program with 2,000+ affiliates that generated $485,000+ in revenue.

Substack Recommendations drove 78% of Lenny’s new subscribers at scale.

Your best customers become your best salespeople…if you make it easy for them. Set up an affiliate program on Gumroad. Ask happy buyers to share. Make the product so good that word-of-mouth does the heavy lifting.

💡You can learn all these inside my AI Digital Product Builder

Step 8 — Scale Into the Ladder

Once you have consistent sales at Rung 1 or 2, it’s time to add the next rung.

This is where the compounding really kicks in.

Not because you worked harder.

Because you built a system that works whether you’re online or not.

The scaling moves that matter most:

✅ Add an upsell immediately after purchase.
Justin Welsh’s one-click upsell became a $22K/month recurring product. When someone buys your $47 guide, show them your $150 course right after. The conversion rate on post-purchase upsells is significantly higher than cold traffic.

✅ Turn your best content into a recurring community.
Once you have 100+ buyers, launch a membership. Even at $15/month, 100 members is $1,500/month in predictable recurring revenue. That changes your entire relationship with money and stress.

✅ Reinvest into free content.
Sahil Bloom’s lesson: treat your free content as a “cost center” that feeds your revenue engine. The more value you give away publicly, the more your list grows, the more your products sell. Free content is not a sacrifice — it’s the top of your funnel.

✅ Build testimonials into your system.
After every sale, follow up and ask for a result. A single specific testimonial (”I made my first $500 online using this guide in two weeks”) is worth more than any paid ad.

The 8-Step Summary

This is the framework. Eight steps, proven pattern, real results from named creators who did exactly this.

But here’s where it gets interesting for you specifically.

Because every single one of these steps…niche research, validation, building, launching, scaling…can now be done faster and leaner with AI.

That’s exactly what we’re covering in the next section. 👇

Share The Digital Creator

The AI Execution Layer: How to Compress Years into Months

Let me tell you about two creators who used AI at very different levels.

⏩ Creator A treats AI like a search engine.
They open ChatGPT, type “write me a blog post about productivity,” copy the output, paste it directly, and wonder why it doesn’t convert.

⏩ Creator B builds an AI system.
They have a Claude Project loaded with their brand voice, their audience profile, their product context. Every workflow is saved. Every output goes through a human layer. They produce 4–5 high-quality pieces of content per week in roughly 2 hours.

Same tool. Completely different results.

The difference isn’t the AI.

The difference is how systematically you use it.

Here’s what the data shows:

Teams using AI for both creation and automation produce 75% more content per week while saving an average of 14 hours. AI-powered creators are capturing a disproportionate share of the creator economy….which hit $253 billion globally in 2025.

But here’s the part that matters most for you right now:

AI doesn’t just compress the creation process.

It compresses every single step of the product ladder we covered.

Research. Validation. Building. Launching. Scaling.

All of it…faster, leaner, with one person doing the work of a whole team.

Let me show you exactly how.

The Solo Creator AI Org Chart

Before we get into specific tools and workflows, here’s the mental model I want you to carry.

Think of your AI stack not as a collection of chatbots…but as a team.

Every great business has roles. As a solo creator, AI fills those roles for you:

Most successful solo founders spend $80–$200/month on this stack.

That’s not an expense. That’s the cost of replacing a team that would otherwise run you $10,000+/month.

AI by Task: The Exact Tool for Every Step of the Ladder

Here’s the part that most creators get wrong.

They pick one AI tool and use it for everything.

That’s like using a hammer for every job in your house…including the ones that need a screwdriver.

Different tasks need different tools.

Here’s the map:

🔍 Step 1 — Niche Research & Validation

→ Perplexity for live web research: what’s trending, what’s selling, what people are asking right now

→ Claude for synthesis: “Here’s what I found…what does this tell me about positioning?”

→ Reddit + Claude for demand scanning: dump relevant Reddit threads into Claude, ask it to extract the top pain points, the exact language people use, and the solutions they’re currently buying

💡 This is literally the process I described inside the AI Digital Product Builder — with live data to validate your idea before building a single slide.

✍️ Step 2 — Content Creation (Free Value + Newsletter)

→ Claude is the best tool for long-form writing, editing, and voice consistency. Hands down.

In head-to-head tests, Claude consistently produces writing that sounds more natural and human than any other AI. Zapier’s 2026 analysis confirmed Claude “thinks strategically and offers options unprompted”...it doesn’t just execute instructions, it anticipates what you actually need.

But the key is building a system, not using one-off prompts.

The 4-layer Claude content system:

Layer 1 — Personal Preferences: Set your universal rules in Claude’s settings. Your default tone, your writing style, what you never want Claude to do. This applies to every conversation automatically.

Layer 2 — Claude Projects: Create one project per content type…newsletter, threads, product content. Each project holds your brand voice guide, your audience persona, your formatting rules. You upload these once. Every conversation in that project stays on-brand automatically.

💡 I’ve covered exactly how to set this up for your newsletter content system here: Newsletter Content System with Claude Skills

Layer 3 — Claude Skills: Build reusable workflow modules for your most repetitive tasks. A Hook Creator skill. A Newsletter Drafter skill. A CTA Writer skill. Each one is a markdown file that encodes your standards, your voice, your output format. You build it once and use it forever.

💡 Start here if you’re new to Skills: The 3 Claude Skills Every One-Person Business Needs — and if you want to understand the difference between Skills and Projects: Claude Skills vs. Claude.md — Context Engineering

Layer 4 — Automation: Connect Claude to your tool stack using n8n or Make. New subscriber joins → welcome sequence triggers.
New newsletter published → social posts auto-generated.
New product sold → delivery email fires.

💡 I built this exact kind of agentic automation workflow here: AI Agentic Workflow for Building a Substack Employee — it shows how to wire Claude into your content pipeline so the system does the distribution work while you focus on creation.

After the initial setup of 15–25 hours, this system runs on 1–2 hours per week.

I covered how to build this entire system in detail here: How I Built a Claude-Powered Operating System

📦 Step 3 — Building the Product

This is where AI becomes a genuine game-changer.

What used to take months now takes a weekend.

Here’s the proof:

Nat Eliason went from tweet to $200,000 in pre-sales for his AI app-building course in one week.
He validated by posting, took pre-orders immediately, then used Cursor to build the actual course content at speed.

Nicolas Cole and Dickie Bush loaded their entire Category Newsletter Creator curriculum — 50,000+ words — into Claude as reusable Skills.
That curriculum has generated over $10 million in revenue. AI didn’t replace their expertise. It organized and delivered it at scale.

If you want to follow along with me, here you go 🚀

🚀 Step 4 — Launching and Selling

→ Claude for sales copy:

Landing page copy, email sequence drafts, launch tweets, and thread hooks…all in your voice, all in one project.
The key is how you prompt it. I broke down the exact prompting system that makes Claude write in your voice (not generic AI output) in The Prompting Playbook from Claude — including the specific structures that work for sales copy and launch sequences.

→ Claude Skills for the Educational Email Course:

Build the nurture sequence once, let Kit deliver it automatically to every new subscriber forever.
If you haven’t built your Skills yet, start here: The 3 Claude Skills Every One-Person Business Needs — it covers exactly which Skills move the needle first, including the ones built specifically for email and content delivery.

If you want to go deeper on how Skills and your newsletter content system connect, this one goes further: Newsletter Content System with Claude Skills.

→ NotebookLM for repurposing:

Drop your newsletter issue or product content into NotebookLM and generate audio overviews, mind maps, and social content variations from a single source.
The full repurposing workflow…how to turn one piece of content into 10 assets across platforms — is inside the 30-Day One-Person Business Playbook with Claude. It includes the exact sequence for building this into your weekly content system so repurposing stops being an afterthought and becomes automatic.

⚙️ Step 5 — Automating Delivery and Support

This is where solo creators build real leverage.

Once your product is live and your email sequences are set up, automation turns your business into a machine that runs while you sleep.

The automation flows that matter most at the start:

✅ New subscriber → welcome email → nurture sequence → product offer

✅ New product purchase → delivery email → thank you → upsell offer

✅ New newsletter published → social posts auto-drafted in Claude → queued in Typefully for review

💡 Watch me doing the automation using Kit here

AI Is the New Standard — Not a Shortcut

Let me be clear about something important.

AI gives you speed.

You give it soul.

The creators who fail with AI are the ones who hand over the thinking. Who treat Claude as a ghostwriter and paste outputs without editing. Who skip the human layer entirely.

Ali Abdaal’s Head of Content said it best:

“AI is really good at giving us adequate stuff. It’s like a cover band that hits all the notes perfectly, but you’re never gonna feel that element of personality, that element of jazz.”

The human 20–30% is what makes content connect.

But that 70–80% that AI handles? Research, structure, formatting, repurposing, optimization, delivery, automation?

That used to take a team.

Now it takes one person with a $100/month stack and a clear system.

That’s not a shortcut. That’s a structural advantage.

And right now — in 2026 — the creators who build that system are going to have a gap between them and everyone else that only grows wider over time.

Don’t get left behind.

The Real Timeline (AI Compresses Execution, Not Trust)

This is the most important thing in this entire section.

Write it down.

Here’s what I mean.

Building a product? AI cuts that from months to a weekend.

Writing your email sequence? AI cuts that from weeks to an afternoon.

Repurposing your content across platforms? AI cuts that from days to an hour.

But the time it takes for a stranger to trust you enough to give you money?

That can’t be automated. It can’t be prompted. It can’t be shortcut.

Lenny Rachitsky wrote weekly for 9 months before he charged a dollar.

Tiago Forte ran 9 cohorts before his revenue became meaningful.

AI didn’t exist when they started. So in that sense, you have a real advantage…the building is faster now.

But the trust still compounds the same way it always has.

Through showing up. Consistently. With genuine value. Over time.

Here’s the realistic timeline to keep in mind:

This isn’t a guarantee. But it is the pattern.

And with AI in your corner, you’re doing this in months what used to take creators years.

The One Thing That Separates Creators Who Make It

I’ve watched a lot of creators go through this journey.

The ones who build real businesses aren’t always the most talented.

They’re not always the best writers.

They’re not always the ones with the biggest networks or the most polished products.

The ones who make it have one thing in common.

They don’t stop before the compounding kicks in.

That’s it.

They post when nobody is watching.

They email a list of 43 people like it’s 43,000.

They build the product even when they’re not sure it’ll sell.

They refine and iterate instead of starting over.

The compounding is real. The ladder works. The framework has been proven by dozens of creators across different niches and platforms.

But it only works if you’re still in the game when it starts to click.

So stay in it.

Build the rung in front of you.

One step at a time. 👇

Here’s Where You Go From Here

We’ve covered a lot of ground today.

Let me bring it all together before you close this tab.

Here’s what you now have that most creators don’t:

✅ The truth about why most digital products earn $0…and what separates the top 1%

✅ A 4-rung product ladder with real creator proof at every level

✅ An 8-step repeatable framework pulled from the people who actually made it

✅ An AI execution layer that compresses months of work into days

✅ A reality filter so you go in with the right expectations…and don’t quit before the compounding kicks in

This is not theory.

Every framework in this issue has been lived by real creators…and tested in my own business.

I went from a relatively unknown creator with a few thousand followers to 75,000+ on X, 18,000+ newsletter subscribers, and 1,000+ digital products sold…with zero paid ads.

Not because I had some secret advantage.

Because I followed a system. Rung by rung.

And now you have that system too.

The question is: what are you going to do with it?

Your Next Step Depends on Where You Are Right Now

I don’t believe in one-size-fits-all CTAs.

So here’s what I’d recommend based on where you are in the journey:

If you’re just starting out and haven’t built your system yet…

Start with the free playbook.

It’s the same lean framework I use…condensed into a guide you can implement this week. It covers how to identify your niche, set up your lead magnet, and get your first piece of content working for you before you spend a single dollar.

→ Grab the Lean AI-Powered Business Playbook for Creators — Free 🚀

If you’re ready to build and launch your first digital product…

This is exactly what the AI Digital Product Builder course is built for.

It walks you through every step…skill discovery, audience validation, product creation, landing page copy, launch sequence, and the evergreen sales system…with AI doing the heavy lifting at every stage.

It’s the most complete resource I’ve built for going from “I have an idea” to “I have a product that sells.”

→ Get the AI Digital Product Builder — $67

If you’re already creating and want to go deeper, faster…

Join as a Premium Member.

Upgrade Now 🚀

For $39/month (or $150/year — that’s less than $3/week), you get:

One Last Thing Before You Go

I want to say something directly.

The creator economy is not slowing down. It’s accelerating.

And the gap between creators who build systems and creators who stay on the treadmill is widening every single month.

AI has made it cheaper, faster, and more accessible than ever to build a real knowledge business.

But it’s also raised the bar on what “good” looks like.

The creators who will win the next 2–3 years are not the ones who just use AI.

They’re the ones who build systems with AI…and then show up with the human layer that no tool can replicate.

Your perspective. Your experience. Your voice.

That’s the asset AI can’t build for you.

Everything else? This playbook has you covered.

Now go build the next rung.

Your next level awaits. Go get it.

Sharyph

P.S. If this issue gave you a useful framework, the single best thing you can do is share it with one creator who needs it. Forward this email. Post about it. Tag me on X @sharyph_. Nothing grows a creator’s business faster than word of mouth from people who got genuine value. And nothing makes me happier than seeing this land in the right hands.

P.P.S. If you’re not yet a Premium Member, the AI Digital Product Builder course alone is worth more than the annual membership price. Join here and get access to everything — including 45% off the course — from day one.

Until next time — Sharyph | Founder of The Digital Creator

Want to put your product or service in front of 18,000+ creators? → Sponsor this newsletter

Most people are still using AI like a search engine with a chat box.

They type a question. They get an answer. They move on.

And if that’s you…you’re not alone. But you are falling behind.

Here’s what nobody tells you:

That’s not a hack. That’s a fundamentally different relationship with the tool.

We are no longer in the era of “AI tips and tricks.” If you’ve been collecting prompts trying to find the magic words, I need you to hear this: the prompt era is over.

We are now in the Agent-Native Era.

Being agent-native means you don’t just use AI…you orchestrate it. You delegate to it. You build systems around it. You think like the CEO of a small AI workforce, not like someone filling in a chatbot form.

The question for 2026 isn’t “are you using AI?”

In this issue, I’m going to break down the 9 skills that separate the people operating at the top 1% of AI capability from everyone else.

This isn’t a list of shiny apps or hidden features.

These are foundational skills, the kind that compound. The kind that, once you build them, make every AI tool you touch 10x more powerful.

Here’s what we’re covering:

1. Natural Language Mastery — English is the new code

2. Designing and Curating AI Skills — Your reusable instruction library

3. Mastering Super App Platforms — Your AI command center

4. Orchestrating Cloud-Based Agents — AI that runs 24/7, even when you’re offline

5. Human Foundations — The real 1% differentiator (it’s not technical)

6. Asynchronous Automation — Building your 24/7 workforce

7. Computer and Browser Use — Agents that take the wheel

8. Token Budgeting and Model Selection — The cost advantage nobody talks about

9. Real-Time Voice — The Jarvis workflow

Let’s get into it.

Skill 1: Natural Language Mastery (English is the new code)

Let me say something that might sting a little.

If your prompts still look like this...

“Act as a senior marketing expert with 10 years of experience. You are professional, insightful, and data-driven. Your task is to...”

...you are using 2023 techniques in a 2026 world.

Those templated prompts were necessary back when AI models needed rigid structure just to understand what you wanted. You had to essentially program them with roles and constraints because they couldn’t infer intent on their own.

That era is gone.

The shift is from prompting to describing.

Andrej Karpathy, one of the original researchers behind modern AI, put it simply: “The hottest new programming language is English.”

He wasn’t being poetic. He was being literal. The person who can articulate what they want with the most precision and context... wins. Every time.

This is what practitioners are now calling Context Engineering.

It’s not about finding magic words or secret formulas. It’s about giving the model enough rich, specific context to do genuinely great work. Think of it less like filling in a template and more like briefing a highly capable team member who just joined your business today.

What does that look like in practice?

Instead of:

“Write a newsletter about AI productivity.”

Try:

“I’m writing a weekly newsletter called The Digital Creator, aimed at solo creators and one-person businesses who are intermediate AI users. They already use Claude but haven’t built any systems around it yet. This issue needs to feel like a breakthrough moment for them... like they just learned something that changes how they see AI entirely. The tone is direct, no fluff, and always actionable. Start with a hook that challenges how they currently think about prompting.”

See the difference?

Same tool. Same model. Completely different output.

The quality of what you get from AI is a direct reflection of the quality of what you put in. Not the cleverness of a prompt hack. Not a special format you copied from a YouTube video. Just clear, specific, contextual description of what you actually want.

This is also why I wrote an entire issue on the Prompting Playbook from Claude earlier this year. Prompting is not dead... it just grew up. It’s no longer about tricks. It’s about communication.

The practical takeaway here is simple.

Before you type your next message to any AI... pause. Ask yourself: “If I were briefing a new hire on this task right now, what would I tell them?” Write that. Send that.

That habit alone will put you ahead of 80% of AI users.

Skill 2: Designing and Curating AI “Skills” (Your reusable instruction library)

Here’s a question worth sitting with.

How many times this week did you write the same type of prompt... for the same type of task?

A newsletter draft. A thread hook. A product description. A content brief.

If you’re doing that repeatedly, you’re not building leverage. You’re just doing manual labor with a fancier keyboard.

This is the problem that AI Skills solve.

Think of it like this. Imagine you hired a brilliant new team member. On their first day, you spent two hours walking them through your brand voice, your content format, your standards, your audience, and your process. They took notes. They understood.

Now imagine you had to give that same two-hour briefing every single morning before they could do anything.

That’s what most people are doing with AI today.

A Skill solves this. You brief the agent once. It learns your standards. And every time you call on it, it already knows what you need.

Here’s the part most people get wrong though.

They think building a Skill means sitting down and writing a long instruction document from scratch. That sounds tedious, so they never do it.

The actual workflow is the opposite of that.

Here’s how I do it and how I teach it:

1. Do the task first. Have a conversation with Claude, write the thing, get the output you want.

2. Refine it. Push back, adjust, iterate until the output is exactly where you want it.

3. Then ask Claude to document it. Literally say: “Based on everything we just did, turn this into a reusable Skill file I can use next time.”

Claude builds the Skill for you. From the work you already did.

That’s it. No blank page. No manual writing. Just capture what worked and bottle it.

I’ve published an entire issue on exactly how this works, including the difference between a Skill file and a Claude.md context file, which are two different things that most people confuse. If you haven’t read it yet, this is the one.

And if you want to go deeper on the technical side of building and organizing these Skills inside Claude Code, this guide walks through the full setup.

The real power of Skills is what happens over time.

Every mistake becomes a rule. Every time an output misses the mark, you update the Skill so it never happens again. The system gets better every week... without you having to think harder or work longer.

Creator documented this beautifully. She built three core Claude Skills covering her audience profile, voice DNA, and business context. Her result? “The first output is already 80% there. I go from draft to almost-final in a fraction of the time.”

That’s the compounding effect of Skills. You invest a little time building them once. And they pay you back on every single task after that.

The future of this is even more interesting.

We are already moving toward agents that auto-update their own Skill files based on your feedback. You correct an output, and the agent notes the correction and updates its own instructions. The system teaches itself... from you.

That’s not science fiction. That’s where this is heading in 2026.

The one-person business that masters Skill design today will have a compounding advantage that’s almost impossible to replicate later.

Subscribe now

Skill 3: Mastering the “Super App” Platforms (Your AI command center)

Here is how most people use AI right now.

They open a chat tab. They ask a question. They wait. They copy the answer. They paste it somewhere else. They close the tab. They open a new tab. They start over.

One task. One tab. One result.

That’s not a workflow. That’s a bottleneck.

The people in the 1% are not working like this. They’ve set up what I call an AI command center. A single, centralized platform where multiple agents run simultaneously, connected to their tools, their files, and their browser... all at once.

This is what the new generation of “super app” AI platforms is designed for.

The two biggest right now are Claude Desktop and Cursor. And as of April 2026, both have been redesigned from the ground up around one core idea: you should not be waiting for one task to finish before starting the next.

Claude Code’s April 2026 desktop redesign shipped parallel sessions, an integrated terminal, file editor, diff viewer, and drag-and-drop panes. The mental model shifted completely: you’re not waiting for one task to finish, you’re running several at once and checking in as results arrive.

You are running several agents at once. And checking in when results arrive.

That is a fundamentally different relationship with AI than what most people have today.

What does this look like in practice?

Imagine you start your morning and you launch Claude Desktop. In one session, you have an agent researching competitors in your niche and pulling together a summary. In another, a second agent is drafting this week’s newsletter based on the Skill you built last month. In a third, another agent is reviewing your last five X posts and identifying which format drove the most engagement.

You go make your coffee.

You come back to three completed outputs.

The built-in browser changes everything too.

One of the most underrated features of these platforms is the built-in browser. Instead of you manually visiting a URL, copying content, pasting it into a chat, and then prompting... the agent just navigates the web itself. It reads the page. It processes the information. It brings back what you need.

All three major platforms... Claude, Codex, and Cursor... are converging on the same vision: one environment where you can chat, do knowledge work, write and run code, and automate workflows, connected to your existing tools, with an in-app browser as the interface layer.

The browser is becoming the agent’s hands.

And here’s how I think about choosing between these platforms, because this question comes up constantly in my inbox.

Claude Desktop / Claude Cowork is the best choice for knowledge work. Writing, research, content strategy, newsletter drafting, product building. If you are a one-person business or a creator, this is your home base. I wrote an entire issue on how I turned Claude into a full operating system for my business earlier this year. That is what is possible when you stop treating it like a chat window and start treating it like infrastructure.

Cursor is built for developers who want an IDE with AI baked in. If you write code regularly, it’s worth knowing. But for the creators and solopreneurs reading this, Claude Desktop is where you want to be.

The practical move right now is this.

Stop opening AI in a browser tab like it’s Google. Download Claude Desktop. Set it up as your morning workspace. Organize your Projects inside it. And start running more than one thing at a time.

Because while everyone else is waiting for one tab to load...

You’ll already be three tasks ahead.

Skill 4: Orchestrating Cloud-Based Agents (AI that never clocks out)

Let me ask you something.

What happens to your business when you close your laptop?

For most people, the answer is: everything stops.

Tasks pause. Research stops. Follow-ups don’t get sent. Content doesn’t get drafted. The business is only as productive as the hours you’re actively sitting in front of a screen.

That was acceptable before agents existed.

It is not acceptable anymore.

This is a fundamentally different category from the AI tools most people are using today. A chat session is something you have to show up for. A cloud agent is something that shows up for you.

Here’s what that looks like in practice.

Right now, you can set up an agent connected to your Telegram. You message it at 7am: “Draft a content brief on the top AI trends from this week, and have a first draft of Thursday’s newsletter ready by 9.” You put your phone down. You go for a walk. You come back to a brief, a draft, and a summary of sources... waiting for you in the same chat.

You didn’t sit at a desk. You didn’t open a browser. You texted your AI like you’d text a team member.

That’s exactly the model behind tools like OpenClaw, which connects AI to your existing messaging apps including WhatsApp, Telegram, Slack, and Discord... running 24/7, even when your laptop is closed. One creator who tested it described the experience simply: it turns rough notes into ready posts and queues them on schedule. No more blank-page mornings.

OpenClaw crossed 100,000 GitHub stars within its first week of launch in early 2026. That’s not hype. That’s developers and creators voting with their attention because they recognized something real: this is what the agent era actually looks like.

And it’s not just third-party tools catching up.

Claude Code now supports native Channels... a plugin-based feature that connects your Claude sessions directly to Telegram, Discord, and iMessage. You send a message from the app, Claude processes it using your local environment, and replies in the same chat.

Think about what that means for a one-person business.

Your AI is now reachable anywhere you already are. Not in a new app you have to remember to open. Not in a dashboard you have to log into. In the apps you’re already checking 30 times a day.

The interaction model stops being “sit at terminal, type a command.” It becomes “text your assistant whenever something comes up.”

This is also where MCP changes everything.

With MCP active, your agent isn’t just answering questions about your calendar. It’s reading your calendar, checking for conflicts, rescheduling meetings, and sending confirmations. It’s not just writing a draft based on your Notion notes. It’s reading your Notion directly, pulling the relevant context, and delivering a finished piece.

Over 50 official MCP connectors now exist for Claude alone... covering Google Workspace, Slack, Notion, HubSpot, Canva, Asana, and more. The agent doesn’t just live inside a chat box. It lives inside your entire tool stack.

The practical setup I recommend for creators and solopreneurs is this:

Start simple. Connect your AI to one messaging channel you already use every day. Telegram is the fastest setup... roughly 5 minutes. Then give your agent one repeating job. A morning content brief. A weekly newsletter research pull. A daily inbox summary.

Let it run for a week.

You will feel the shift immediately. Because for the first time, your business will be doing work while you’re not working.

That’s not a productivity hack. That’s leverage.

And once you feel what it’s like to wake up to work already done... you won’t go back to the one-tab, one-task, one-result model. Ever.

Skill 5: The “Human Foundations” (The real 1% differentiator)

Here’s the part nobody wants to say out loud.

All four skills we’ve covered so far... the natural language mastery, the Skills library, the super app platform, the cloud agents... none of them will save you if this one is missing.

And this one isn’t technical at all.

The single most important skill for operating in the top 1% of AI users in 2026 is a fundamentally human one.

It’s the ability to delegate well.

Let me explain what I mean. Because most people misunderstand this.

When I talk about delegation in the context of AI, I’m not just talking about handing off tasks. I’m talking about three specific human capabilities that directly determine the quality of everything your agents produce.

1. Communication

The agent can only work with what you give it. If you can’t articulate what you want with clarity and specificity... the output will be average. Every time. No model upgrade in the world fixes a vague brief.

This is why the best AI users are almost always the best communicators. Not the most technical people. Not the ones with the most tools. The ones who can describe a desired outcome so precisely that even a new team member would know exactly what to produce.

2. Taste

Here’s the uncomfortable truth about AI output.

It will always produce something. It will rarely produce something great without a human who knows the difference.

That’s what taste is. It’s the ability to look at a piece of work and know whether it’s good enough... or whether it misses the mark and why.

And taste can’t be prompted. It can’t be automated. It comes entirely from domain expertise and experience.

If you’ve spent five years building an audience on X, you have taste for what makes a thread perform. If you’ve written 60 newsletters, you know when a hook lands and when it doesn’t. If you’ve sold over $30,000 in digital products... you know exactly what a compelling offer sounds like versus one that goes quiet on launch day.

Your AI doesn’t have that knowledge. You do.

That expertise is the ceiling of what your agents can produce.

Raise your expertise, and you raise your AI output. It compounds in both directions.

3. The Managerial Mindset

Think about the best manager you’ve ever seen or worked with.

They didn’t do everything themselves. They didn’t micromanage every decision. But they knew the work well enough to set clear expectations, spot when something was off, and give precise feedback that actually improved the output.

That’s exactly the mindset you need with AI agents.

If you’re a good manager of people, you will be a good manager of agents. The skills translate almost perfectly. Setting context. Defining deliverables. Reviewing critically. Iterating with direction rather than just saying “make it better.”

This is also why I built the AI Digital Product Builder course the way I did. The entire framework is built around one principle: your AI-powered business is only as strong as the human judgment running it. The tools are the easy part. Knowing what to build, who to build it for, and whether what you’ve built is actually good... that’s the skill that creates the separation.

The course walks you through that judgment development step by step. From identifying your monetizable skill... to building and validating your offer... to using AI to create, launch, and sell it. All 8 weeks of it.

If you’re ready to stop experimenting and start building something real with AI, the AI Digital Product Builder is $67 and it’s the most complete system I’ve built.

Back to the point…

Studies now consistently show that top AI performers don’t just fire off more prompts. They spend significantly more time on the front end... defining what they want. And more time on the back end... reviewing, refining, and redirecting output.

They are deeply involved. Just at the strategic level, not the execution level.

That’s the distinction most people miss.

They either do everything themselves and burn out. Or they hand everything to AI and wonder why the output is flat and generic.

That’s not a software feature. That’s a human skill.

And it’s one you can start building today.

Share The Digital Creator

Skill 6: Asynchronous Automation (Your 24/7 workforce)

I want you to think about something for a second.

How many tasks in your business happen at the same time every week?

Researching content ideas. Pulling your newsletter brief together. Summarizing your inbox. Repurposing last week’s post into threads and LinkedIn content. Sending follow-up emails after someone downloads your lead magnet.

Same tasks. Same sequence. Every single week.

You do them manually. They take hours. And because they take hours, you either rush them, skip them, or burn out trying to stay consistent.

Here’s the thing.

And in 2026, setting that automation up no longer requires you to know how to code. It doesn’t even require you to drag and drop blocks in a visual builder for 45 minutes.

It requires you to describe what you want in plain English.

That’s it.

This is what I mean by asynchronous automation. You build the system once, in your own words. And then it runs... on a schedule... in the background... while you focus on the work that actually needs your brain.

Let me show you exactly how this works.

You open Claude with n8n connected via MCP. You type:

“Build a workflow. Every morning at 8am, pull the latest item from my RSS feed, summarize it into 3 key points, then rewrite it as a LinkedIn post, an X thread, and a 1000-word newsletter paragraph. Queue each one to the right platform.”

Claude reads that. Asks one clarifying question. Then builds the entire workflow... node by node... directly inside your n8n instance. No manual configuration. No dragging. No debugging for an hour.

n8n’s own team confirmed that the first generation of a workflow with Claude now takes roughly three minutes instead of the usual 10 to 30 minutes it takes to build one manually. What used to require knowing node types, connection logic, and configuration options now just requires you to describe what you want.

That is a fundamental shift in who can build automation.

Before, automation was for developers. Now it’s for anyone who can explain a task clearly.

And this is where it gets really interesting for creators.

The workflows that move the needle most for a one-person business are not complex. They are simple, repeating pipelines that currently eat hours every week.

Here are three I would build first if I were starting from scratch today:

The Content Repurposing Pipeline. One newsletter issue goes in. Out comes an X thread, a LinkedIn post, a Substack Note, and three hook variations. Scheduled automatically. One source, four outputs. This alone gives you back 3 to 5 hours per week.

The Morning Intelligence Brief. A scheduled trigger fires at 7am. It pulls your top newsletter metrics from the past 7 days, checks for new replies, scans a curated RSS feed of your niche, and delivers a clean summary to your Telegram. You start every day already informed, before you open a single tab.

The Lead Magnet Follow-Up Sequence. Someone downloads your free playbook. A trigger fires automatically. Claude drafts a personalized follow-up based on which resource they downloaded. It queues in Kit. You review and approve in 60 seconds. Done.

These are not complex systems. They are simple workflows... described in plain English... running quietly in the background while you do the work only you can do.

I’ve covered the agentic workflow side of this in depth in a previous issue that walks you through the exact setup. If you haven’t read it, that’s a good next step after this one.

The bottom line is this.

Every hour you spend doing a repeating task manually is an hour you are choosing not to build leverage. Asynchronous automation is how you get the same output... with a fraction of the ongoing input.

Build it once. Let it run. Move on to the next thing.-world

Skill 7: Computer and Browser Use (Agents that take the wheel)

Picture this.

You’re on your phone between meetings. You need a pitch deck exported as a PDF and attached to a calendar invite before the call starts in 12 minutes.

Old world: you rush to your laptop, open the file, export it, find the invite, attach it, pray you didn’t make a typo in the filename.

New world: you message Claude from your phone. “Export the pitch deck as a PDF and attach it to today’s 2pm meeting invite.” You put your phone back in your pocket.

By the time you sit down... it’s done.

This is not a concept. This is what Claude Computer Use, launched in March 2026, actually does. Claude opens your apps, navigates your browser, fills in forms, clicks buttons, moves files... all by itself. It sees your screen the same way you do. And it acts on what it sees.

Anthropic described it simply when they launched it: “Anything you’d do sitting at your desk.”

Anything you’d do sitting at your desk.

That is an extraordinary statement. And it’s one that most people haven’t fully absorbed yet.

Here’s how it works under the hood.

Claude Computer Use runs on a continuous loop. It takes a screenshot of your current screen. It analyzes what it sees... the buttons, the text, the state of the application. It decides the next action. It executes it. It takes another screenshot to verify it worked. Then it repeats.

There’s no API required. No pre-built connector. No special integration. If a task exists on your screen... Claude can do it. Because it’s working exactly the way you work. Eyes on the screen, hands on the keyboard.

This closes a gap that has frustrated AI users since the beginning.

Up until now, AI could only help you with tools that had connectors or APIs. If your tool wasn’t integrated... you were on your own. Computer Use removes that wall entirely. If it lives on your screen... Claude can touch it.

The priority order matters here though.

When Claude executes a task in Cowork, it follows a smart hierarchy. First, it checks whether a connector exists for the tool you need... Gmail, Google Drive, Notion, Slack. If a connector is available, it uses that. Faster. More reliable. If no connector exists, it pivots to Chrome, navigating the browser directly to get the job done.

This is the right approach. Connectors when possible. Browser when necessary. Full computer use as the last resort for anything that requires real desktop interaction.

Where this is most useful right now:

For creators and one-person businesses, the highest-value use cases are the ones that currently require too many manual steps to be worth your time:

Filling out a form on a platform that has no API. Navigating a CMS to format and publish a draft. Pulling data from a dashboard that doesn’t export cleanly. Cross-referencing information across two or three open tabs. Logging into a tool, running a report, downloading it, and saving it to the right folder.

All of that is fair game now.

The honest caveat.

This is still early. Anthropic said so themselves when they launched it. The vision-action loop is slower than API-based automation. Complex tasks with lots of steps can still trip up. And as of March 2026, it’s macOS only... with Windows support confirmed but not yet shipped.

The right mental model is this: use Computer Use for supervised tasks, not fully autonomous ones. Set it up, watch it run the first few times, build trust gradually. Don’t hand it access to your bank account or your email on day one.

But for the right tasks... the ones that are repetitive, multi-step, and tedious... it is genuinely one of the most impressive things I have seen AI do in the last two years.

The gap between “cool demo” and “actually useful daily tool” closed in March 2026.

And most people haven’t noticed yet.

Skill 8: Token Budgeting and Model Selection (The cost advantage nobody talks about)

Here’s a cost most people are paying without realizing it.

Technically capable. Massively overkill. And getting more expensive every month.

This is the reality of AI in 2026 that almost no one is talking about in creator circles. Frontier models... the most powerful, most capable AI available... are not getting cheaper. They are getting more expensive as labs add capability and raise prices simultaneously.

GPT-5.5 on OpenRouter currently runs at $5 per million input tokens and $30 per million output tokens. Meanwhile, open-source models like DeepSeek V4 run agentic tasks at $0.094 per million tokens. That is a 30x price differential for tasks that often produce indistinguishable output.

This is the skill nobody teaches: using the right model for the right task.

Not every task deserves your most expensive model. In fact, most tasks don’t. And once you internalize this, you stop paying frontier prices for everything and start building a model stack that is both more powerful and dramatically cheaper.

Here’s a simple mental framework I use:

Frontier models for high-stakes, complex reasoning.

Strategy. Multi-step problem solving. Nuanced creative work. Anything where the quality of thinking directly determines the quality of the output. Claude Opus, GPT-5.5... these are tools for your hardest problems. Not your routine ones.

Mid-tier models for standard content work.

Drafting newsletters. Writing threads. Summarizing research. Repurposing content. Claude Sonnet handles almost all of this at a fraction of Opus pricing. A 1,500-word blog post via Claude Sonnet costs roughly $0.02 to $0.03. That is not a typo. Two to three cents.

Open-source models for high-volume, low-complexity tasks.

Classifying emails. Summarizing documents. Generating outlines. Bulk processing. DeepSeek V4, Llama, Qwen... these models handle routine production work at near-zero cost and they are better than GPT-4 was just two years ago.

This is where OpenRouter changes the game entirely.

OpenRouter is a single API that gives you access to over 300 models from Anthropic, OpenAI, Google, Meta, DeepSeek, Mistral, and dozens more... all from one interface. No separate accounts. No switching tools. No rebuilding your workflow.

You pick the model. You set cost limits. You route different tasks to different models automatically based on what makes sense.

OpenRouter recently raised $113 million in funding and now processes 25 trillion tokens per week... a fivefold increase from just six months ago. That growth tells you everything about where serious AI users are heading. They are not defaulting to one expensive model for everything. They are routing intelligently.

The practical setup for a one-person business looks like this:

The people doing this well are spending $30 to $50 per month total on AI inference while running workflows that would cost $300 to $500 per month if they used frontier models for everything.

That’s not a small difference. That’s the difference between AI being an expense and AI being leverage.

One more thing worth knowing.

The open-source models available today are not the limited, inferior alternatives they were in 2023. DeepSeek, Qwen, Kimi, and GLM are on serious shortlists for agentic workflows... not because they’re cheap, but because for many tasks they perform at or near frontier quality. The gap is closing fast.

Knowing which model to use for which task... that is a compounding skill. Every month you get better at it, your output improves and your costs go down at the same time.

That’s a rare combination. Build it.

Skill 9: Real-Time Voice and the “Jarvis” Workflow (The interface shift nobody sees coming)

Think about every Iron Man movie you’ve ever watched.

Tony Stark doesn’t sit at a desk typing prompts into a chat window.

He talks. His AI listens. Things happen.

That scene... the one where he’s walking around his workshop, speaking out loud, and his entire digital world responds in real time... used to be pure fiction.

It’s not fiction anymore.

Real-time voice AI is the final interface shift. And most people reading this are still ignoring it completely because they think it’s just a gimmick. A slightly hands-free version of typing.

It’s not. It’s a fundamentally different way of working.

Here’s the gap that most people don’t see.

You type at roughly 40 to 50 words per minute. You speak at 120 to 150 words per minute. That’s a 3x speed difference before we’ve even talked about what you can do with your hands while your mouth is working.

When your interface is voice... you are not sitting at a desk. You are walking. Cooking. Commuting. Exercising. Moving through your day. And your AI is keeping up with you the entire time.

That’s not a minor productivity upgrade. That’s a different relationship with your work.

The landscape in 2026.

Several platforms are now competing seriously in this space.

ChatGPT Voice with GPT-4o is currently the benchmark for conversation quality. Near-instant responses, emotional awareness, handles interruptions naturally. The voice doesn’t feel robotic. It feels like a conversation. For brainstorming, thinking out loud, and working through creative problems verbally... this is the strongest option available right now.

Gemini Live from Google made a significant move by making real-time voice free and globally available. It also adds camera and screen-sharing capabilities... meaning your AI can see what you’re looking at while you talk to it. You point your phone at something and ask questions. That is a multimodal shift that matters.

Claude’s voice mode is available on mobile and now includes Claude Code voice mode specifically for developers who want to speak instructions to their coding agent in the terminal. Claude powered by ElevenLabs technology means the voice quality is noticeably natural compared to earlier iterations.

All three are moving fast. Every month the latency drops, the quality improves, and the capabilities expand.

But here’s where it gets genuinely powerful for one-person businesses.

Voice isn’t just for answering questions. When you combine real-time voice with the computer use capability we covered in Skill 7... you get something that is very close to the Jarvis experience.

You speak. Your AI listens. And then it acts on your computer.

“Open Notion and create a new project page for the product launch.”

“Draft a reply to the last email from my subscriber and save it as a draft.”

“Pull last week’s newsletter metrics and give me a verbal summary while I make coffee.”

You’re not sitting at a keyboard. You’re not copying and pasting. You’re not switching between tabs. You’re just... talking. And your digital workspace is responding.

The most productive professionals in 2026 are not choosing between AI and voice. They’re using both simultaneously. They dictate into Claude. They let Claude respond by voice. They continue the conversation while their hands do something else entirely.

The practical entry point right now.

If you’re on iOS or Android, start simple. Open the Claude app or ChatGPT and switch to voice mode. Take a task you’d normally type... a content brief, a thread outline, a product idea... and just talk through it instead.

Don’t type. Don’t plan what you’ll say. Just speak like you’re explaining it to a smart colleague over coffee.

You’ll be surprised how fast the ideas move when you remove the keyboard from the equation.

Then notice what changes. Your thinking becomes more natural. Your briefs become richer because you ramble in useful ways when you speak. And the AI keeps up... asking clarifying questions, pushing back, building on what you said.

That’s the beginning of the Jarvis workflow.

And the gap between where we are today and the full vision... an AI that listens, acts, executes, and reports back entirely through voice... is closing faster than most people realize.

By the end of 2026, not having a voice workflow will feel like not having a smartphone in 2012.

You’ll still function. But you’ll be visibly, measurably slower than the people who figured it out early.

Your Roadmap to the 1%

Let’s bring this home.

9 Skills. That’s what separates the top 1% of AI users from everyone else in 2026.

Not nine apps. Not nine subscriptions. Not nine prompt hacks copied from a YouTube video.

Nine foundational skills that compound into a genuine operating advantage.

Let me give you the one-sentence version of each:

✅ Natural Language Mastery ... the person who describes what they want most clearly wins every time.

✅ AI Skills Design ... stop re-explaining yourself. Build once, deploy forever.

✅ Super App Mastery ... stop using AI in a browser tab. Build a command center.

✅ Cloud Agent Orchestration ... your business should be working when you’re not.

✅ Human Foundations ... your expertise, taste, and delegation ability are the ceiling of your AI output.

✅ Asynchronous Automation ... willpower is not a business strategy. Systems are.

✅ Computer and Browser Use ... if it lives on your screen, your agent can do it.

✅ Token Budgeting ... pay frontier prices only for frontier problems.

✅ Real-Time Voice ... the interface is shifting. Get ahead of it now.

Notice something?

The first four are about building the infrastructure. The fifth is about developing the human judgment to run it. The last four are about deploying it at maximum leverage.

That’s the complete picture of what agent-native looks like in 2026.

Not someone who uses more AI tools. Someone who thinks differently about AI altogether.

They don’t ask: “How can I use this tool?”

They ask: “What does this agent need from me to do this job well?”

That shift... from user to orchestrator... is the whole game.

The one thing to do this week.

Don’t try to implement all nine skills at once. That’s how you end up overwhelmed and back to your old workflow by Thursday.

Pick one skill. Just one.

If you’re still explaining yourself from scratch every time you open Claude... start with Skill 2. Build your first Skill file this week.

If you’re still running one AI tab at a time... start with Skill 3. Download Claude Desktop today and set up three Projects.

If you have repeating tasks eating your mornings... start with Skill 6. Map out one automation in plain English and set it up in n8n.

One skill. One week. One system running that didn’t exist before.

That’s how the 1% is actually built. Not in a weekend sprint. In small, compounding moves made consistently over time.

Want a more in-depth and practical route?

If you want to skip the trial and error and get the actual tools I use to run my business on these 9 skills... that’s exactly what the Premium Membership is built for.

For $39/month (or $150/year, which saves you 40%), you get:

✅ Premium Claude Skills ... ready-to-use instruction files for the exact recurring tasks creators face every week. Hook writing, newsletter drafting, content repurposing, SEO optimization, and more. Built and tested. Plug in and run.

✅ AI Agents Collection ... a curated library of agent setups you can deploy for your business without building from scratch.

✅ Substack SEO/AEO Implementation Toolkit ... the system I use to optimize every issue for both search and AI-powered discovery.

✅ 45% off all paid courses ... including the AI Digital Product Builder.

✅ Community access ... a group of builders who are actually implementing, not just consuming content.

The goal of the membership is simple. You read the newsletter and get the thinking. You join the membership and get the tools.

Join as a Premium Member here →

The transition from AI user to AI orchestrator doesn’t happen all at once.

It happens one skill at a time. One system at a time. One task you never have to do manually again.

But here’s what I know after building this for two years and helping thousands of creators do the same.

Once you feel what it’s like to wake up to work already done... once you feel your output double while your working hours stay the same... once you realize your AI team is quietly compounding in the background while you focus on the things only you can do...

You don’t go back.

That’s not hype. That’s just what systems do.

Now go build yours.

Until next time,

Sharyph
Founder, The Digital Creator

You have a call in 45 minutes.

You open your calendar. You see the name. You vaguely remember the last conversation. You open a new tab. Then another. Then LinkedIn. Then their website. Then you realize you’re still not sure what you actually want to get out of this meeting.

This is how most us operates today…

And here’s what I know about most creators and solopreneurs: they’re good at showing up. They show up consistently. They take the calls.

But they walk into those calls unprepared.

Not because they don’t care. Because prep takes time they don’t have. So they wing it and they walk away from meetings thinking, “I should have asked that,” or “I forgot to bring up the thing that actually mattered.”

That’s not a discipline problem. That’s a system problem.

And in 2026, a system problem has a system solution.

Here’s what I built and what you’re going to copy today.

A Meeting Prep Agent inside Claude. No automation tools. No code. No new subscriptions.

Just a Claude Project, a smart setup, and one copy-paste prompt and before every meeting, you get a clean briefing document that tells you exactly who you’re meeting, what to cover, what to ask, and what outcome to drive toward.

I’m going to walk you through the exact build, step by step.

By the end of this issue, you’ll have a working agent. Not a concept. Not “something to try someday.”

A working agent.

Let’s build it.

What This Agent Actually Does

Before we build, let me show you what you’re actually building.

Because most creators, when they hear “AI agent,” picture something complex. Something technical. Something that requires a developer or a 3-hour YouTube tutorial.

This isn’t that.

What you’re building is essentially a smart briefing assistant that lives inside Claude…and knows how to prepare you for any meeting, any time, in under 2 minutes.

Here’s what it produces every single time you use it:

A ready-to-read Meeting Brief that covers:

1. Who you’re meeting Not just a name. Context. What they do, what they’re building, what matters to them…pulled from whatever you feed it (email threads, notes, a LinkedIn bio, a proposal).

2. The purpose of this specific meeting What does success look like at the end of this call? Your agent frames the meeting around an outcome — not just a topic.

3. Talking points tailored to them Not generic conversation starters. Points that are relevant to this person, this conversation, this moment.

4. Smart questions to ask The kind of questions that make you sound like you’ve done your homework…because you have. Except Claude did it for you.

5. Potential friction points or objections If there’s something sensitive, unresolved, or worth navigating carefully…your agent flags it. So you’re not caught off guard.

You get all of that… in under 2 minutes…before every single meeting. Just by pasting a prompt.

That’s what “AI as a co-pilot” actually looks like in practice. Not hype. Not vague promises about productivity. A real output you can read over coffee before you jump on a call.

Why Claude? Why not a dedicated tool?

There are tools built specifically for meeting prep. Lindy, Reclaim, Otter…they all have versions of this. And some of them are genuinely good.

But here’s why I’m showing you the Claude-native approach first:

You already have it. If you’re reading this newsletter, there’s a strong chance Claude is already part of your workflow. No new subscription. No new login.

It’s faster to set up. We’re talking 10 minutes to a working agent…not a multi-step integration with your calendar, CRM, and email.

You control the output. Dedicated tools give you their format. With Claude, the brief looks exactly the way you want it to look. You define the structure. You define the tone. You own the system.

And honestly? The output quality is exceptional…when you set it up correctly. Which is exactly what this issue is about.

What you’ll need before we build:

• A Claude account (free works, Pro is better…more on why in a moment)

• 10 minutes

• One upcoming meeting to test it on

That’s it.

Let’s get into the build.

The Setup: Build Your Meeting Prep Agent in Claude

This is where we build.

Follow these steps in order. Don’t skip ahead. The magic isn’t in any single step…it’s in how they work together.

Before the steps, I want to show how the output looks. Here is the meeting prep. I got to meet with (this is a made-up scenario)

Let’s start creating this agent…

Step 1: Create a Dedicated Claude Project

If you’ve been using Claude as a regular chat window…opening a new conversation every time…you’ve been leaving a lot of power on the table.

Claude Projects are different.

A Project gives Claude a persistent memory of who you are, what you do, and how you want it to work…across every conversation inside that project. You set it up once. It remembers everything. Every time you come back, it picks up right where you left off.

For a Meeting Prep Agent, this is everything. Because your agent needs to know you — your business, your clients, your context — before it can prep you for anything.

Here’s how to create one:

1. Open Claude at claude.ai

2. On the left sidebar, click “Projects”

3. Click “New Project”

4. Name it something clear — “Meeting Prep Agent” works perfectly

5. You’re in. Now let’s configure it.

Step 2: Write Your System Prompt (The Agent’s Brain)

This is the most important step.

Your system prompt is the set of instructions Claude follows every single time you use this project. It’s what transforms a general-purpose AI into your meeting prep agent…one that understands your context, your style, and your standards.

Think of it as hiring an assistant and writing their job description. The more specific you are, the better they perform.

Read more

Let me ask you something honest.

How many times have you fixed a prompt… and broken it again two weeks later?

Maybe you added a new instruction. Maybe you migrated to a newer Claude model. Maybe you just changed one line — and suddenly the output you depended on stopped making sense.

You didn’t change your goal. You didn’t change your business. You just changed one thing.

And now it doesn’t work.

Here’s what that chaos actually looks like:

• You paste in the same prompt you’ve been using for months

• The output is vague, too long, or misses the point entirely

• You add more instructions to “fix” it

• Now it’s over-explained and the new instructions conflict with the old ones

• You delete things. Add things. Move things around.

• One hour later, you’re not sure what’s in the prompt or why

And the worst part?

You have no way to know if the change you made actually helped — or just shifted the problem somewhere else.

You’re not engineering. You’re guessing.

I’ve been there.

When I first started building Claude workflows for The Digital Creator — my content research agent, my newsletter Skills, my offer analysis system — I made every mistake in the book.

I’d pile instructions on top of instructions. I’d tell Claude to “always do X correctly” without actually giving it the capability to do X. I’d patch one failure and create three new ones.

The result was a graveyard of half-working prompts and a lot of wasted hours.

Here’s what I’ve learned after building AI systems inside a real creator business:

Prompting isn’t guessing. It’s engineering. And engineering has a process.

The creators who get consistent, reliable output from Claude aren’t smarter. They’re not using secret prompts.

They’re using a system.

Last month, Anthropic released a session from their internal engineering team called The Prompting Playbook. It was a live demo by a real applied AI engineer showing exactly how they debug prompts, fix failure modes, and build new agents from scratch.

I watched it, took it apart, and rebuilt it as a framework you can actually apply…whether you’re fixing a broken Claude Skill or building a new AI agent for your business.

By the end of this issue, you’ll have:

✅ A 3-step process for cleaning up any prompt before you touch the logic

✅ The 3 most common Claude failure modes — and the exact fix for each one

✅ A proven loop for building new AI agents that work reliably from day one

✅ 5 rules you can apply to every prompt you write going forward

This isn’t theory. This is the exact playbook Anthropic engineers use…applied to the kind of systems you and I are actually building.

Let’s get into it.

The Two Scenarios You’ll Always Face

Every prompting problem you’ll ever have falls into one of two buckets.

Scenario A:

You have an existing prompt that used to work — and now it doesn’t.

Maybe you migrated to a newer Claude model. Maybe someone on your team edited it. Maybe your use case grew and the prompt didn’t grow with it. Whatever the reason, the output has degraded and you’re not sure why.

Scenario B:

You’re building something new from scratch.

A new agent. A new Claude Skill. A new workflow you’ve never built before. You have a goal, a blank page, and no idea where to start.

Both scenarios feel completely different.

But here’s what most people miss:

Both scenarios need the exact same foundation before you do anything else.

That foundation is evaluations.

You Can’t Fix What You Can’t Measure

Here’s the trap most creators fall into.

They change something in their prompt. The next output looks better. They move on.

Two days later, it breaks again in a different way. They fix that. Something else breaks.

They’re not improving the prompt. They’re playing whack-a-mole.

The reason this keeps happening is simple: they have no way to measure whether a change actually worked.

An evaluation — or “eval” — is just a set of test cases you run your prompt against every time you make a change.

Think of it like this.

You wouldn’t redesign your newsletter layout and just assume it improved open rates. You’d check the numbers. Before and after.

Prompts work the same way. Every time you change something, you need a before and after. Without that, you’re not iterating — you’re drifting.

What Goes Into a Good Eval

You don’t need 100 test cases. You need the right test cases.

Anthropic engineers use three types:

1. Control cases — The easy ones. Unambiguous questions your prompt should always get right. These make sure your changes didn’t break what was already working.

2. Edge cases — The tricky ones. Scenarios where the model has failed before. By including these, you’re making sure old failure modes don’t quietly slip back in.

3. Boundary cases — The limit cases. Situations where Claude needs to know what’s outside its scope. Where it should hand off to a human, refuse a request, or escalate instead of guessing.

For most creator workflows…a newsletter Skill, a content research agent, an offer analysis system…five to ten test cases is enough to get real signal.

Why This Matters For Your Claude Skills

If you’ve been building Claude Skills or Claude Code workflows for your business, this is the piece most people skip.

They build the Skill. It works on the first try. They ship it.

Then three weeks later…after a model update, after adding new instructions, after the use case evolved slightly…it starts producing inconsistent output. And because there’s no eval, there’s no systematic way to diagnose what changed.

Before you touch your next prompt, write down three to five test cases first.

What should it always get right? What has it gotten wrong before? Where should it stop and ask instead of guess?