You can tell when something came straight out of a blank AI chat window. It has a smell.

It starts with the em dashes. One sentence uses three of them, every one doing a job a comma could have handled.

Then the vocabulary. Everything is “without the overwhelm” now. Every lesson is a “game changer”. Every point is a “hard truth”, and at least once we are all going to delve. Nobody has delved out loud since 1840, but the robot delves with its whole heart.

Then the quirky sentence structures. “It’s not a newsletter, it’s a movement.” A generic opener like “Most creators” or “In today’s fast-paced digital world,” which is a sentence no living person has ever said to another living person. And the headers, each one wearing a tiny rocket ship (why?) 🚀

You probably scrolled past multiple posts like this today and chances are, you don’t remember one of them.

You retained none of it because the writing came from nobody.

Here’s what happens when you hand a machine a blank slate and ask it to fill the space: it grabs the most average version of every word. It produces the most likely sentence based on a mathematical formula, then the next most likely one, all the way down the page. It reads fine and mean nothing at the same time.

The first mistake is relying on AI for your ideas. AI isn’t where your ideas come from. It’s where they get bigger. You bring the opinion, the stories only you have, the stance that you believe in, the expertise and origin story that is uniquely yours and it helps you shape that into content that is unmistakably written by you.

Give it your real material and what comes out could only have come from you. Give it nothing and it gives you the generic nobody.

And Yes, I use AI to produce my content because it saves time, but I don’t let AI run the show - I’m the producer, director and the actor reading the script I wrote.

Today I’m sharing my process that anybody can use to create better, more specific, relatable content that is authentically YOU.

Here’s my process: Before I enlist AI’s help with a social post, newsletter, video script, or any other content I’m teaching, I open Wispr Flow and I talk.

Why out loud: talking tends to surface speech patterns that typed responses hide. Your go-to phrases, words you use in unique ways, your sentence rhythm, and the way you explain things. That's the raw material your Voiceprint uses, and you produce it naturally when you speak.

This isn’t tidy talking. I pace around my office and ramble at the screen for ten minutes, full of ums, half-finished sentences, a couple of tangents, and the occasional “no wait, back up.”

Wispr Flow catches all of it in the AI chat window. What lands on the screen is an ugly wall of text nobody but me could decode, and that in that mess is the gold. It’s a download of my thoughts before I’ve cleaned them up for company. Only then do I let AI put the pile in order.

The brain dump handles the raw material and I could stop there. But there is a second piece to this that will result in an even better, more YOU final product and I see people miss this one often.

You have to train your AI properly.

Your AI has to know your style, your preferences, beliefs, anlgles and everything about you and your creator life so it can write like you.

It’s important that it know the words you prefer and the words you would never be caught using (I keep a list, and “delve” sits at the top), as well as the takes you’ll defend and the trends you skipped. Give it none of that and you’re back to the average of everyone. Give it a real picture of how you think, speak, and process information and it stops guessing.

People skip this because it sounds like work, and because nobody handed them a simple way to start.

So I built one especially for creators, by a creator who writes for a living and got tired of sounding like a generic press release.

It’s called Your AI Voiceprint. A free prompt (well 2 actually) that turns any AI into a relentless interviewer. It asks how you write, what you stand for, and what you’d never touch, then compiles your answers into one file. The file is your voiceprint and you can feed that file to Claude, ChatGPT, or whatever AI tool you use.

You will spend about 30 minutes answering the question, preferably with Wispr Flow or any other Voice-to-text tool, answering out loud, and the file is yours to sharpen and update as circumstances change.

It won’t write your work for you and that’s the point. It makes sure that when AI helps, the words still sound like a unique individual - a.k.a. You.

Grab it here:

GET YOUR AI VOICEPRINT PROMPT

Other ways I can help you:

→ Creator Cash Flow Club — weekly playbooks, Substack 360 course, monthly group coaching, Creator Tool Vault, and priority DM support.

→ 9-To-Thrive Accelerator — eight weeks of 1:1 personalized coaching tailored to exactly where you are in your journey.

→ Substack Newsletter Audit — find out exactly what’s costing you subscribers and revenue, with specific fixes you can implement right away.

→ Substack Setup Sprint — four focused sessions to get your Substack set up the right way: positioning, homepage, offers, and a simple growth plan.

→ 60-Minute Clarity Call — one call to untangle your strategy and walk away with a clear direction.

It was Saturday morning. Second cup of coffee. Content calendar open for May. Twelve posts to plan, three offers in rotation, and one question I had been avoiding for six months: which of my last 90 days of posts had actually driven a paid upgrade, and which had been idling.

Open rates were not the answer. Two of my highest-open-rate posts that quarter pulled zero upgrades. Meanwhile, a 700-word post I almost killed because I thought it was too short — drove twelve.

I had been flying on instinct. Some weeks I picked a banger. Some weeks I picked a dud. And every dud was another seven days I published the wrong thing while three offers sat unsold.

The dashboard was not going to tell me. Substack will give you opens, clicks, maybe a cute little line graph. It will not tell you which post structure made someone upgrade. That is an analysis problem. Analysis problems are what AI is built for.

I needed a better method, so I built the workflow I now run every month. I call it the Receipts Audit Method. It took 45 minutes to set up, runs in 12 minutes once it is calibrated, and the first time I ran it on my last 90 days, it traced $5,850 in upgrades to a single post inside a 72-hour window — and flipped three assumptions I had been operating on all year.

Below I’m sharing the prompt, the setup, and what came back.

Why dashboard analytics fail solo creators

Dashboard analytics tell you which posts got opens. They do not tell you which posts drove paid revenue.

Substack will show you that a post got 41% opens. It will not tell you that of the seventeen people who upgraded that month, fourteen of them clicked through from the post that got 28% opens. The conversion data and the engagement data live in different views. Most creators look at the prettier one and write more posts that look like the one with the line going up.

That is how you end up with a content plan optimized for opens and a revenue line that does not move.

The fix is not “post more.” The fix is to merge data sets and ask a model to find the pattern. A correlation across 90 days of writing is too weak for the human eye and too obvious for Claude to miss.

Keep your publishing streak intact

The One-Hour Newsletter System produces a complete issue — researched, written, optimized, image prompt, A/B subject lines, and scheduled in Substack — in under an hour. 12 AI skill files. The exact workflow behind every issue you read here.

A repeatable system for quickly creating a newsletter in your voice →

GET THE SYSTEM

Here is the Five Steps Receipts Audit Method.

Below the line: the exact setup steps, a copyable prompt, the four real patterns the audit returned on my own archive, and the anti-pattern I had been writing for years. This one is so good that it’s available only to my paid subscribers.

Read more

The worst part was always when I’d get fed up with the draft and type something into Claude — “write me a paid issue about welcome sequences” — and watch it return four paragraphs that opened with “in today’s evolving creator economy landscape.”

I’d delete it, open a blank doc, and start over.

That was every Sunday for my very first attempts at using AI to write my newsletter.

Most newsletter operators hit that moment and decide AI doesn’t work for newsletters. They’re wrong, and the reason they’re wrong is the prompt.

Writing a newsletter involves multiple steps. Trying to cram it all into one prompt is like putting eight jobs into one job description: topic idea, research, outline, headline, body, CTA, SEO, image, Notes, LinkedIn Post, carousel, etc.

Each one has different inputs and different rules. Collapse all eight into a single “write me a newsletter” prompt, and the AI does the only thing it can with no instructions. It averages. It returns whatever sits in the statistical middle of every type of content it was trained on, and that’s the source of the generic-creator-content sound. The AI can do the job if it’s well equipped, but if the production architecture isn’t correct, the end product will be mediocre at best.

So I stopped asking AI to do eight jobs at once and built a system that runs each one separately, in order, with rules that match my voice, expertise, business, proprietary methods, and offers. I drop a topic in, and the system gives me back the full package: body, headlines, SEO, image brief, repurposing copy, all of it. Ready to schedule.

Four hours dropped to one. I publish multiple formats a week to 8,600+ subscribers and I haven’t lost another Sunday since.

I have evolved my process of using AI for my newsletter over the last 12 months, and so has Claude. I started before Claude cowork, code and skills were even born Claude has evloved and as writers, we need need to pivot.

My system is executed in Claude cowork, but it can also be done in chat or code. The key to all of it is developing the appropriate skills to match each task so that the finished product delivers it exactly as you intend it to.

It’s so good now that I bundled the skills I use every week with a playbook and a couple of foundational prompts, all of which are now included in The One Hour Newsletter System.

If your Sundays still look like mine used to, this is the fix.

If you’ve been telling yourself you’ll get to AI workflows when you have time, the system is the time.

— Carrie

Other ways I can help you:

→ Creator Cash Flow Club — weekly playbooks, Substack 360 course, monthly group coaching, Creator Tool Vault, and priority DM support.

→ 9-To-Thrive Accelerator — eight weeks of 1:1 personalized coaching tailored to exactly where you are in your journey.

→ Substack Newsletter Audit — find out exactly what’s costing you subscribers and revenue, with specific fixes you can implement right away.

→ Substack Setup Sprint — four focused sessions to get your Substack set up the right way: positioning, homepage, offers, and a simple growth plan.

→ 60-Minute Clarity Call — one call to untangle your strategy and walk away with a clear direction.

The Prompt-to-Live Workflow is a four-step process for shipping a fully designed, interactive landing page in Lovable in under a day using Perplexity for pre-planning, Lovable’s planning feature for structure, single-ask prompt iteration for polish, and a CRM bridge for lead capture.

My client called Thursday. He needed a landing page live by Monday morning.

Not a simple page. A lead magnet opt-in page with an embedded interactive assessment widget I’d built separately. A form that feeds their CRM. And their team had to approve the everything before we shipped.

Monday was four business days away, minus the approval loop. Call it three days of real build time.

If I handed this to a designer and a developer working together, we’d ship by next Thursday, if we rushed it. A week too late.

I run two different CRM platforms (HubSpot and Go High Level) for different parts of my business. On a normal timeline, I’d build this inside one of them. I use them for everything from landing pages to email sequences to pipeline management.

A form on a landing page built inside the CRM means the lead lands in the same database running the rest of the business. Sequences fire. Tags apply. Pipeline advances. No duct tape, no middleware.

But building a polished landing page inside most CRMs, wiring the follow-up sequence, and getting a custom assessment widget to embed cleanly? That’s a week of work on a generous timeline. I had three days.

So I opened Lovable.

Two prompts. Three hours. The page was staged with the assessment widget embedded, the form capturing, the brand typography matching. I sent the staging link to the client at lunch. They approved everything by 4 p.m. I pointed the domain. Live before dinner.

Here’s how I did it, and the one thing nobody tells you about Lovable that almost wrecked the project at the finish line.

Why I use Perplexity before I open Lovable

This is the step most people skip, and it’s the reason their Lovable builds take 15 prompts instead of 3.

Before I opened Lovable, I spent about 45 minutes inside Perplexity working through the page specifications. Perplexity is my thinking partner for any build. I walked through the page architecture, the audience, the sections, the copy tone, the embed requirements, the form fields, and the visual direction. I asked it to challenge my assumptions. It pushed back on a few section decisions and caught a sequencing issue I would’ve missed.

By the time I opened Lovable, I wasn’t thinking anymore. I was executing. I had a clear brief with every decision made.

This is why the entire Lovable build took two prompts and three hours instead of two days of back-and-forth iteration. The specs were already airtight.

If you’re opening Lovable cold with a vague idea of what you want, you’re going to burn credits fixing things that should’ve been decided before you wrote a single prompt. Perplexity first. Lovable second. Every time.

What is the Prompt-to-Live Workflow?

The Prompt-to-Live Workflow is a four-step process for shipping a fully designed, interactive landing page in under a day. It works for projects where the timeline is too tight for a traditional CRM build or a designer-developer handoff.

Four steps. Roughly three hours of focused Lovable work. One live page. Here’s the full breakdown.

Step 1: Plan inside Lovable

Lovable has a built-in planning feature now. Use it first, every time.

You describe the page: audience, outcome, sections, copy tone. Lovable produces a structured plan with section names and a short description of each. You approve, edit, or redirect before it writes a line of code.

This is where your Perplexity pre-work pays off. When your specifications are already tight, the Lovable plan lands close to final on the first pass. You’re confirming, not brainstorming.

Skip the planning step and Lovable will build you something. It won’t build you the right thing.

Step 2: Upload a model page

Screenshot or link a page that looks like what you want. Upload it to the project. Tell Lovable to use it as the visual reference.

On this client’s page, the first attempt missed. Hero was off, typography was wrong, button treatment didn’t match. I went back twice. By round three, the layout was right. Budget two to three rounds on this step if the reference is specific.

Step 3: Iterate with single-ask prompts

Do not prompt Lovable in paragraphs. Prompt in single asks.

“Change the hero headline to [text].” “Move the assessment widget above the lead form.” “Make the primary button orange (#FF6B35), not pink.”

Every targeted prompt burns one or two credits. Every vague, multi-paragraph prompt burns five and produces something you have to fix. One ask at a time. This is not a suggestion.

Step 4: Get the form data back into the CRM

This is where the project almost fell apart.

Lovable builds beautiful pages. It does not natively connect to a CRM. Which means the form on your shiny new page is capturing data into nowhere useful unless you build the bridge yourself.

I asked Lovable directly about the options. There are five. The right one depends on the existing stack and the complexity of the data you’re capturing.

Five ways to get Lovable form data into a CRM

This is the part that separates a staged page from a live lead-gen asset. I’m listing all five because you need to pick the right one before you start building, not after.

1. Embed a native CRM form. Most CRMs provide an embeddable form, either an iframe or a JS snippet. Leads land directly in the CRM with full attribution and workflows already wired up. Zero backend work.

The trade: styling is constrained. HubSpot allows CSS overrides; others are more rigid. The embedded form won’t perfectly match your custom design. And it only works for the lead capture form, not for passing assessment or quiz data.

2. Direct CRM API integration. Keep your custom-designed forms and have a backend function call the CRM’s API on submit. Full design control. Works for both forms and assessments because you can push quiz answers as custom contact properties.

The trade: each CRM has its own API quirks, and you’re locked to that CRM until you rewrite the integration. HubSpot has a built-in Lovable connector that makes this cleanest.

3. Zapier, Make, or n8n automation. Form submits to a webhook, the automation platform catches it, routes it to the CRM and anywhere else you want (Slack, Google Sheets, email notification).

The trade: ongoing subscription cost, a third-party dependency, and a small delay.

4. Email notification only. Form submits, someone gets an email, they enter it manually. Dead simple. Does not scale. Leads get missed. Use this as a temporary backup, never the primary method.

5. Database log + CSV export. Store every submission in a Lovable Cloud database table. Export to CSV and import into the CRM periodically. A safety net. Not real-time. Layer this under one of the above if you want a backup of every lead regardless.

What I did: The client uses HubSpot. HubSpot’s native embed form allowed enough CSS styling overrides to match the Lovable page cleanly. I dropped in the embed, the form captured directly into HubSpot, and the existing workflows fired on submit. Twenty minutes from embed to confirmed test lead. Zero Zapier. Zero backend code. Problem solved.

If the client had been on a CRM with rigid embed styling, I’d have gone with option 2, the API integration through a Lovable Cloud Edge Function. One extra hour of setup, full design control, and the ability to push assessment data as custom contact properties.

When NOT to use the Prompt-to-Live Workflow

Three situations where you should build inside the CRM instead.

The client already runs their entire business on one CRM platform and every lead needs to feed the system that handles their email, pipeline, and automations natively. Don’t add integration complexity to a stack that doesn’t need it.

The page is part of an existing funnel with lifecycle stages, tagging rules, and drip sequences already built. The CRM handles that. Lovable does not.

Read more

Reclaim 5 hours every week

The One-Hour Newsletter System produces a complete issue in your voice — researched, written, optimized, image prompt, A/B subject lines, and QA tested in in less than 60 minutes. Build your newsletter system once and create unlimited newsletters. The exact workflow behind every issue you read here. →

I NEED THIS!

It is Monday morning. I am sitting at my desk with coffee (of course) and I have to produce a paid newsletter, a free newsletter, a LinkedIn carousel, and three sales emails by Friday. A year ago, that would have meant 25 hours of writing, editing, second-guessing, and rebuilding the same template from scratch every single time.

Instead, I open Claude Cowork, type one sentence, and watch a skill take over. Headline, subtitle, full body, SEO package, image prompts, CTA block. Done in 12 minutes. Two more skills handle the carousel and the emails.

By 11 a.m., I am done. The rest of the day is for client calls, my course, and grilling out with my family on the back deck.

This is what Claude Skills do for solo creators. They turn the messy, repeatable parts of running a business into one-click workflows that match your exact voice, your exact offers, your exact systems.

If you have ever copied and pasted the same prompt into ChatGPT for the 40th time, watched your “draft” come out generic and corporate, or wasted an afternoon getting Claude to remember how you write. Skills are the fix.

Here is what this guide covers:

• What Claude Skills are and why they matter for solo creators

• The difference between Claude Chat, Claude Cowork, and Claude Code

• How Claude Skills work in each product

• 10 real creator use cases (so you know what is possible)

• How to build your first skill with Claude Skill Builder in 30 minutes

• What goes inside a SKILL.md file

• Where to store your skills and how to organize them

• How to install and run a skill (with the trigger trick most people miss)

• Where to find pre-built skills on GitHub and how to vet them safely

• Claude Skills vs. Claude Projects: when to use each

• The five mistakes that kill skills and how to fix them

• The honest math on whether skills are worth the time

• A 5-step plan to get your first skill live this week

Let’s go.

What are Claude Skills, and why do they matter for solo creators?

Claude Skills are folders of instructions, examples, and rules that teach Claude how to do a specific task the way you want it done, every single time. Each skill is a reusable workflow you build once and run forever. For solo creators, that means your newsletter, your sales emails, your LinkedIn carousels, and your offer pages all spin off the same brain you trained, instead of starting from a blank prompt.

A skill is a small file inside inside a Skill folder, called SKILL.md. That file holds the rules: what task this skill performs, when Claude should use it, and how Claude should handle the work step by step.

Anthropic launched Skills in October 2025. Since then, the use case for solo creators has gone from “interesting” to non-negotiable.

Here is why.

A lot of creators are paying $20 to $200 a month for AI tools they barely use. They open ChatGPT, paste a half-baked prompt, and get a draft that sounds like every other Substack post on the internet.

The fix is not a longer prompt. The fix is a skill.

When I write a paid newsletter issue, I do not type instructions into Claude. I open Cowork, point at the topic, and a skill called 9tothrive-paid-newsletter runs the entire workflow. SEO block, headline variants, full body in my voice, image prompts, sign-off, P.S. line.

I have 36 of these in my Skills folder right now. Each one runs a different part of my business.

Skills matter for creators because:

• They make your AI output sound like you, not the average internet writer

• They cut a four-hour task to 12 minutes without losing quality

• They keep your brand voice consistent across every asset, every week

• They let you scale without hiring

• They store your best thinking outside your head

• They know your frameworks, methods, terminology, and everything about you, your business, and your offers

This is the core of running a portfolio business with AI as your unfair advantage.

If you want the complete playbook on how to automate your newsletter, get my One Hour Newsletter System. It’s the complete system for creating your newsletter in under one hour. Includes 12 skill files, 2 prompts and the step-by-step implementation guide. .

What’s the difference between Claude Chat, Claude Cowork, and Claude Code?

Claude Chat (claude.ai) is the conversational version you use in your browser for thinking, writing, and quick research. Claude Cowork is a desktop app for non-developers that runs multi-step tasks across your files, folders, and connected tools. Claude Code is a command-line tool for developers building software. All three run skills. Cowork is the one most solo creators want.

Three products. One brain. Wildly different jobs.

Claude Chat (claude.ai)

This is the version most people meet first. You open a tab, type a question, get an answer. Free for basic use, $20 a month for Claude Pro, and higher tiers if you need long context windows or extra usage.

Use it for: brainstorming, drafting, quick research, asking questions, working through messy ideas. Best for “thinking with Claude.”

Limitation: it cannot reach into your files, run scripts, or execute multi-step workflows on its own.

Claude Cowork

This is the desktop app. You install it on your Mac or PC, point it at a folder, and Claude can read your files, write files, run skills, and orchestrate full workflows without you touching code.

This is where my entire business runs. Every newsletter, every carousel, every sales email starts here.

Use it for producing real deliverables: Word docs, PDFs, spreadsheets, decks. Running multi-step skills. Working across your local files and connected apps like Notion, Gmail, Google Drive, and Substack.

Claude Code

This is for developers. It lives in your terminal. It writes code, edits codebases, runs tests, and ships pull requests. If you do not write software for a living, you do not need it.

That said: many of the best skill templates live inside the Claude Code ecosystem on GitHub. So even if you are not a developer, you will draw on that world as you start to build skills.

The short version: solo creators want Cowork. Some get away with Chat plus a few skills. Almost no creator needs Code.

If you are running a portfolio business solo, Cowork is the unfair advantage.

How do Claude Skills work in each Claude product?

Skills work the same way across all three Claude products. A folder with a SKILL.md file gets stored where Claude can find it, and Claude pulls in the skill when your request matches its trigger description. The difference is in setup. Chat uses uploaded skills via the web interface, Cowork reads from a local Skills folder you point it at, and Code reads from .claude/skills inside your project.

Same skill file. Different storage location. Same end result.

Skills in Claude Chat

Inside claude.ai, you upload skills through the Customize menu under Skills. You zip your skill folder, click the plus button, and Claude installs it inside your account. Once installed, Claude reads the description and triggers the skill when your prompt matches.

This is fine for one or two personal skills. It is not great if you want to maintain 36 of them or share them with a team.

Skills in Claude Cowork

Cowork reads skills from a local folder on your computer. You connect a folder once. After that, every skill inside that folder is available the moment you start a session.

This is what I use. My folder lives at Desktop/Claude Cowork/Skills/. I have 36 markdown files in there. When I tell Cowork to write a paid newsletter, it picks up 9tothrive-paid-newsletter.md and runs.

The benefit: you can edit a skill in any text editor, save it, and Cowork reads the new version next time you trigger it. No upload step. No version drift.

Skills in Claude Code

Skills live inside .claude/skills/ in your project folder. Each skill is its own subfolder with a SKILL.md file. Code uses these to handle repetitive engineering tasks like running tests, refactoring, or generating components in your stack’s exact pattern.

For creators, the takeaway is simple. Pick one product, store your skills there, and write the skill once. The same SKILL.md file will run in any Claude product you connect it to. You build it once, you own it forever.

What can you do with Claude Skills? (10 real creator use cases)

You can build a Claude Skill for any repeatable task in your business. The most useful ones for creators are content production (newsletters, social posts, sales emails), brand management (voice rules, banned words, image specs), audience research (ICP profiles, competitor scans), and offer building (sales pages, lead magnets, course outlines). One creator can run a full content business off ten well-built skills.

These are the ten skills I use most often. Each one runs a workflow I used to do manually for years.

1. Weekly free and paid newsletters. produces the full weeky issue with SEO block, body, image prompts, sign-off

2. About Page/section. For Substack and LinkedIn

3. Hero post writer. long-form pillar for your publication hero section

4. LinkedIn single post. turns any newsletter into a 250-word LinkedIn post

5. LinkedIn carousel. turns any topic into a 7 to 10 slide branded swipe deck

6. Email. Builds email sequences using a fixed architecture

7. Sales page. produces a full long-form sales page from an offer brief

8. Lead magnet. creates the freebie, the landing page, and the welcome sequence

9. Brand voice enforcement. checks every draft for tone, banned words, AI tells, and language style

10. Customer research. pulls real reader pain points from sources you specify

Below that list, I have skills for cold email, copy editing, Notes, images, paid ads, SEO audits, schema markup, customer onboarding, and more. The full Skills folder is the operating system of my business.

The pattern is the same every time: identify a task you do at least once a month, write down how you do it, turn that into a SKILL.md file, save it, point Claude at it.

Skills are the unfair advantage. The average creator is still copy-pasting prompts.

How to use Claude Skill Builder to create your first skill

Open Claude (Chat, Cowork, or Code), type “create a new skill,” and follow the conversational prompts. Claude Skill Builder asks what task you want to automate, walks you through the trigger description, the step-by-step instructions, and the example outputs, then generates a working SKILL.md file you can save into your Skills folder. Total time for your first skill: 15 to 30 minutes.

Anthropic ships a built-in skill called skill-creator (often called Claude Skill Builder). It writes skills for you through conversation. You do not need to know markdown, YAML, or any technical syntax.

Here is the exact process I use to build a new skill in under 20 minutes.

Step 1: Open Claude and trigger the skill. In Cowork, type “I want to create a new skill” or “use the skill-creator skill.” Claude pulls up the skill-creator and starts the conversation.

Step 2: Describe the task. Tell Claude what you want the skill to do. Be specific. “Write a weekly paid newsletter” is too vague. “Write a 1,500-word paid newsletter for my Tuesday issue, with an SEO package, three subject lines, an image prompt, and a P.S. line, in my voice” is the level of detail you want.

Step 3: Describe the trigger. This is the description Claude reads to decide when to fire the skill. Include the exact phrases you will use. “Write my Tuesday newsletter,” “draft this week’s paid issue,” “create a paid newsletter on this topic.” List them all.

Step 4: Walk through the steps. Tell Claude how you want the work done. What goes first? What goes last? What rules matter? What banned words exist? Claude turns your answers into structured instructions.

Step 5: Add an example. Paste in one or two real examples of what good output looks like. This is the single biggest quality lever. The skill will pattern-match against your examples.

Step 6: Test it. Save the skill, restart Claude, and trigger it on a real task. The first version will need tweaks. The third version will be 90% there.

This is also the safest way to build a skill if you are a non-developer. You are not writing code. You are having a conversation. Claude writes the file. If you are still in the early days of building your newsletter, How to Start a Newsletter Business on Substack is the place to begin. Then build your first skill once you have a real workflow to automate.

For the official documentation, Anthropic’s Agent Skills overview goes deep on the technical side.

What goes inside a Claude Skill file?

A Claude Skill file is a single markdown file called SKILL.md. It opens with YAML frontmatter (a small block at the top with name and description). The body holds the written instructions, rules, examples, and any reference materials Claude should follow. The description is the most important part. It is what Claude reads to decide whether to use the skill.

The structure is simpler than it sounds. Here is the anatomy of every skill on my drive.

1. The frontmatter (top 5 lines). A small block between three dashes. It looks like this:

---
name: my-skill-name
description: When to trigger this skill, with exact phrases the user will type
---

The name is the file name. The description tells Claude when to fire. Keep it specific and trigger-rich.

2. The role. A short paragraph telling Claude who it is for this task. “You are a long-form sales page writer for a creator selling a $597 course.” This sets context.

3. The rules. Bulleted list of things Claude must always do. And things Claude must never do. Banned words. Required sections. Voice constraints. Format requirements.

4. The steps. The exact sequence Claude follows when the skill runs. Step one, step two, step three. Numbered. Specific. I ask Claude to ask me questions at the beginning of every skill to ensure we are in alignment.

5. Examples. Real before-and-after snippets showing what good output looks like. The more specific, the better. You can also load your proprietary frameworks, processes or systems for the skill to follow.

6. The output format. Tell Claude exactly how the deliverable should be structured. What sections, in what order, in what file format.

That is it. Six parts. Most of my skills are between 300 and 800 words. The longest are 2,000.

The mistake new builders make is overthinking the file. A 200-word skill with one good example will perform better than a 2,000-word skill with no examples every time. Specifics over volume. Always.

Where to store your Claude Skills (and why folder structure matters)

Store your Claude Skills in one dedicated folder on your computer (a Skills subfolder inside your primary working directory works well). Keep each skill as its own .md file or its own subfolder if it has reference materials. Name the folder Skills, point Claude Cowork at it, and back it up to cloud storage. Your folder structure is your operating system. Treat it like one.

Skills live in three possible places. Pick one and stay consistent.

1. A local Skills folder (recommended for Cowork users). Mine is at Desktop/Claude Cowork/Skills/. Every skill is a single .md file in that folder. When I open a Cowork session and point at the parent folder, every skill is instantly available.

This is the cleanest setup for solo creators. You can see every skill at a glance, edit any one of them in any text editor, and back the entire folder up to iCloud or Google Drive.

2. The Anthropic skills GitHub repository (for shared or open-source skills). Anthropic maintains a public skills repo on GitHub. Some creators store their skills inside a GitHub folder so they can version-control them and share with collaborators. Useful if you have a team. Overkill if you are solo.

3. Inside your Claude account (for Chat). If you only use claude.ai, your skills live inside your account, uploaded through the Customize menu. Easy, but harder to back up and edit at scale.

Folder structure rules I follow

• One folder. Not five scattered across your drive.

• Lowercase, hyphenated file names: paid-newsletter.md, not Paid Newsletter Skill v3 FINAL.md.

• One skill per file. Do not bundle three skills into one giant file.

• A README at the top of the Skills folder listing every skill and what it does.

Backup matters. Your skills are the operating system of your business. If you lose them, you lose months of refined thinking. The same way your Substack Notes strategy is only as strong as the system behind it, your skills are only as strong as the folder they live in.

How to install and run a Claude Skill

To install a skill in Claude Cowork, drop the skill file into your connected Skills folder. Cowork reads it the next time you start a session. To install a skill in claude.ai, zip the skill folder, go to Customize > Skills, click the plus button, and upload it. Once installed, you trigger the skill by typing a phrase that matches its description, or by saying “use the [skill name] skill.”

Installation is the part that scares most non-developers. It should not.

Cowork installation (under 60 seconds):

1. Open the Skills folder you connected to Cowork.

2. Drop your .md file inside it.

3. Start a new Cowork session.

4. Type a phrase that matches the skill’s description.

5. Claude finds the skill and runs it.

That is it. No upload. No restart. No technical setup.

Claude Chat installation (3 to 5 minutes):

1. Put your skill folder on your desktop.

2. Right-click > Compress > zip file.

3. Open claude.ai > Customize > Skills.

4. Click the plus button > Create skill > upload zip.

5. Wait for the install confirmation.

For the full step-by-step from Anthropic, the Claude help article on creating custom skills walks through both paths.

Running a skill once it’s installed

You do not have to call the skill by name. Claude reads the description and triggers the right skill automatically when your request matches.

That said: if Claude is not picking up the right skill, you can force it by typing “use the [skill-name] skill” or “trigger [skill-name].” This is the override.

How to know your skill is running

Look for the skill loading message at the start of Claude’s response. In Cowork it shows a small banner. In Chat it appears as a system message. If you do not see a confirmation that the skill loaded, the skill is not running and you are getting a generic Claude response.

This is the single most common mistake new users make. They think the skill is running. It is not. Then they wonder why the output is generic. Always check the load confirmation.

Where to find Claude Skills you can use as a starting point

The fastest way to start with Claude Skills is to fork an existing one and edit it for your business. Anthropic maintains a public skills repository on GitHub at github.com/anthropics/skills with dozens of example skills you can study and adapt. You can also ask Claude to search GitHub for community skills using a prompt like “search GitHub for public Claude skills related to [your topic].” Always read the full skill before you run it.

You do not have to build every skill from scratch. The Claude ecosystem already has hundreds of pre-built skills you can adapt.

Where to look first

• Anthropic’s public skills repo at github.com/anthropics/skills. this is the official source. Skills here are vetted, documented, and safe.

• The Claude Skill Builder built into Claude itself. ask “show me the example skills available” inside Claude and it will surface the ones already on your machine.

• GitHub search. type “claude skills” or “SKILL.md” into GitHub search. You will find hundreds of community-built skills, ranging from excellent to broken.

• Creator marketplaces. some creators sell skill packs through Gumroad and Substack. These are pre-built for specific use cases like newsletters, sales pages, or cold email.

How to find skills on GitHub without leaving Claude

Tell Claude this:

Search GitHub for public Claude skills related to [your topic]. List the top 10 by stars, with the repo URL and a one-sentence summary of what each does.

Claude will run the search and return a curated list. Pick one, read the SKILL.md file in full, and decide if it is worth adapting.

One word of caution

Read every skill before you run it. Top to bottom. Twice if it has multiple files.

A poorly written skill can produce content that does not sound like you, contradicts your brand rules, produces results that don’t align with your beliefs and methods, or wastes hours of your time. A maliciously written skill can do worse. Execute code, send data, or modify files in ways you did not intend.

Treat every skill from the internet the way you would treat a contract. Read it. Test it on a low-stakes project first. Confirm it does what you want before you run it on a real deliverable.

The good skills will save you hundreds of hours. The bad ones can cause havoc.

Claude Skills vs. Claude Projects: which one should you use?

Claude Projects are containers that hold context, files, and a custom system prompt for an ongoing area of work, useful for “this is my newsletter project” or “this is my book.” Claude Skills are reusable, repeatable workflows that run a specific task, useful for “write my Tuesday newsletter every week.” Most solo creators want both. Use a Project as your home base. Run Skills inside it.

This is the question I get most often from creators new to Claude.

Claude Projects in plain language

A Project is a workspace. You set it up once. You give it custom instructions (”you are my newsletter editor; here is my brand voice”). You upload reference files (your style guide, your customer avatars, your offer catalog). Then every conversation inside that Project automatically inherits the context.

Use a Project when: the work is ongoing, the context is heavy, and you want continuity across sessions.

Best for: book writing, course creation, brand management, ongoing client work.

Claude Skills in plain language

A Skill is a workflow. It runs a specific task. It produces a specific output. It does not need to live inside a Project to work.

Use a Skill when: the task is repeatable, the output is consistent, and you want the same result every time.

Best for: weekly newsletters, sales emails, social posts, lead magnets, sales pages.

How to run them together

This is the setup that pays off for solo creators. Make a Project for your business. Add your brand voice document, your offer catalog, and your customer avatars to the Project knowledge. Then run your Skills inside that Project.

Result: every Skill runs with the full context of your business already loaded. The output sounds more like you, more on-strategy, more on-offer.

Without Projects, every Skill starts from zero context. With Projects, every Skill starts with your full operating manual already in memory.

Common mistakes that kill your Claude Skills (and why testing matters)

The five mistakes that kill Claude Skills are: vague descriptions that do not trigger, missing examples, skills that try to do too much in one file, no testing before production use, and copy-pasted skills from the internet that conflict with your brand voice. The fix is the same in every case. Write specific descriptions. Include real examples. Scope skills tightly to one task. Test on a low-stakes project before you ship.

The skills that fail share the same patterns.

1. Vague descriptions. If your description says “writes content,” Claude will trigger this skill on every prompt that mentions writing. The skill will fire when you do not want it. Or worse, it will not fire when you do. Be specific. Include exact trigger phrases.

2. No examples. A skill without examples is a skill that produces generic output. Examples are how the skill learns your voice. If you skip them, you are rolling the dice every time.

3. Too much in one file. Skills that try to write a newsletter, design the carousel, create the email, and build the sales page all in one file produce average work on every output. Split them. One skill, one job.

4. No testing. You wrote the skill. It looks great. You ship it on a real launch. The output is wrong, the brand voice is off, and you spent two hours rewriting everything by hand. Always test on a low-stakes project first.

5. Copy-pasted skills from the internet. A lot of public skills are written by developers for developers. The voice is wrong. The structure is wrong. The output is wrong for your business. Adapt before you use.

The fix for every mistake is the same: be specific. Use real examples. Scope tightly. Test before you ship.

How I test a new skill

I run it three times on real work. Once on an old asset I already shipped (so I can compare). Once on a current asset I am working on. Once on a deliverable I would be happy to throw away if the skill produces garbage.

If all three pass my review and deliver the result I want, the skill goes into production. If two out of three pass, I revise. If only one passes, I scrap it.

Are Claude Skills worth the time if you’re a solo creator?

Yes, if you do any of the same tasks more than once a month. The break-even on a well-built skill is usually three to four uses. After that, every run is pure time saved. The creators who get the most value from Claude Skills are not technical. They are systems thinkers running content businesses where the same workflows repeat every week.

Here is the honest math.

Building a skill takes 15 to 60 minutes the first time. Writing the trigger, the rules, the steps, and the examples. Testing it. Tweaking it.

Once it is built, every run takes 10 to 20 minutes (depending on how complex the deliverable is) instead of two to four hours of manual work.

If you ship one paid newsletter a week, you save roughly 8 to 12 hours a month per skill. Over a year, that is 100+ hours per skill.

I run 36 skills for different tasks. The math works out to multiple full-time hires worth of saved time per year. And the output is more consistent than what I produced manually three years ago.

The question is not “is it worth it.” The question is “which workflow do you build first.”

The skills with the highest ROI for creators

In order:

1. Whatever you do every single week (newsletter, podcast intro, weekly recap)

2. Whatever you do every single launch (sales emails, sales page, launch sequence)

3. Whatever you do every single client call (call prep, follow-up, proposal)

4. Whatever you do every single month (monthly review, performance report, content audit)

Build the weekly skills first. Build the launch skills second. Build the client skills third. Build the monthly skills last.

If you want to see this in action, every workflow, every skill, every income stream, that is exactly what we run inside the Creator Cash Flow Club. Members get the skills I use, the prompts I run them with, and the playbook for stacking them into a portfolio business.

How to start using Claude Skills this week

The fastest way to start is to pick one task you do every week, open Claude, and tell Skill Builder to create a skill for that task. A solo creator can have their first working skill live in 30 minutes. You do not need a developer, a budget, or a technical background. You need one repeatable workflow, 30 minutes, and the willingness to test it twice.

Do not try to build 36 skills your first week. Build one.

Pick the task you do most often. The one you are sick of doing manually. The one where you feel yourself opening the same template, copying the same prompts, fighting the same blank-page dread.

Your 5-step plan for this week

1. Today: open Claude (Cowork if you have it; Chat if you don’t). Type “I want to create a new skill” and answer the questions. Build a draft skill in 30 minutes.

2. Tomorrow: test the skill on a real piece of work. Compare the output to what you produced manually last time. Note where the output is great. Note where it falls short.

3. Day three: revise the skill. Add a missing example. Tighten a vague rule. Add a banned-word list. Run it again.

4. Day four: ship something with the skill. A newsletter, an email, a social post, whatever the skill is for. Use the output (with light edits) on a real deliverable.

5. Day five: decide if the skill earned its keep. If yes, keep building skill #2 next week. If not, revise once more or scrap and rebuild.

The creators who win with AI are not the ones who buy the most tools. They are the ones who build the smallest set of skills they use every week.

One skill. This week. Then build the next one.

If you want my exact skill stack (the 36 files I run my business on), that is what is inside the Creator Cash Flow Club. New skills drop every month.

The Voice-to-PDF Workflow is a three-input method for building a branded PDF digital product with Perplexity Computer: a voice brain dump captured in Wispr Flow, two to three reference documents that anchor the creator’s voice, and one Perplexity Computer project that produces a finished, branded PDF ready to sell.

I Was Team Gamma For Years. Perplexity Computer Changed My Mind.

Last month I sat down to build The Substack Newsletter Pricing Playbook. I’d been walking creators through pricing their paid tier for 10 months. I knew the content cold. The decision I couldn’t make was how to build it as a product.

Gamma was the default. I’d used it before for pdfs and slide decks for years. I opened a blank project, but something felt off. Every Gamma PDF I’d shipped before looked good on export, but the minute I wanted to edit a section the whole design shuffled. For a product I planned to update every six months, that was a tax I wasn’t willing to pay.

Then I tried Perplexity Chat, which is great for research. I started drafting the playbook in chat. By the fourth message the document was a mess. Generic headers. difficulty nailing brand colors. It read like a 2019 white paper.

The stakes were not abstract. I wanted my Newsletter Pricing Playbook live on Gumroad within two days. There’s a market for it. If I didn’t get it listed this week, it would sit in the queue, and the next three-hour product opportunity behind it would too.

That’s when I opened Perplexity Computer.

Less than two hours later I had a branded, formatted PDF ready for the Gumroad upload page. The playbook is live here.

Why Gamma and Perplexity Chat both fall short for branded PDF products

Gamma builds polished slide decks at speed, but as a PDF product tool, it renders layouts that look great on the first export and fight you every time after. Change one brand color or move a callout and three other sections rearrange themselves. For a product you’ll revise, that eats time you don’t have.

Perplexity Chat is for research, outlines and briefs, not design and layout. Paste a Perplexity Chat output into Word, and you’re looking at unformatted prose with citations. Useful for input, not a finished product.

Perplexity Computer is the third option most creators haven’t tried yet. It launched in February 2026. It orchestrates 19 frontier models — Claude Opus, GPT, Gemini, others — in a single cloud-based agent. You give it a brief and it breaks the work into subtasks, runs them in parallel, and hands back a finished document with layout and branded colors baked in.

Gamma starts with the template. Perplexity Chat starts with the text. Perplexity Computer starts with the product.

What is the Voice-to-PDF Workflow?

The Voice-to-PDF Workflow is a three-input method for building a branded PDF digital product with Perplexity Computer: a voice brain dump captured in Wispr Flow, two to three reference documents that anchor your voice, and one Perplexity Computer project that turns the inputs into a finished, branded PDF you can sell.

Three inputs. One output. No designer, no Gamma reshuffling, no fourth round of revisions.

How to set up the Voice-to-PDF Workflow in five steps

Most people do this on a Saturday morning.

Step 1: Record your brain dump. Open Wispr Flow. Hit record. Talk for five to seven minutes about the product you’re building. Save the transcript as a .txt file.

Step 2: Pull your reference documents. Locate two to three pieces from your existing work — it might be a newsletter post, a framework doc, a web page, a podcast transcript, or a client case study. All in your voice. Export as text files or PDFs.

Step 3: Start a new Perplexity Computer project. Name it after the product or the broader catgory - i.e. digital products. Upload the brain dump transcript and the reference documents. Add a file to the project with your brand colors and fonts.

Step 4: Run the prompt. Paste the Copyable Prompt below. Let Computer draft the outline. Approve or edit the outline before it produces the fully designed document — this is the single highest-leverage moment in the whole workflow.

Step 5: Export, upload, sell. Download the final PDF. Save to Google Drive. Ask perplexity to write the sales page, short description and product delivery page content for your digital product. You can also ask it to create the thumbnail preview and product images (it uses Gemini 3.1 for image generation). Upload everything to Gumroad and hit Publish.

You’ve got the full workflow. You could reverse-engineer a prompt from what I’ve laid out above and probably get a decent result.

Or you could copy the exact prompt I’ve refined over six products and $4,800+ in Gumroad sales — the one that tells Perplexity Computer how to match your voice, structure your sections, render your brand colors, and pause for your approval before it builds anything.

That prompt, plus the week-by-week action steps to get your first product listed, is below for paid subscribers.

Subscribe now

Read more

Here’s Every Tool I Use, What I Pay, and What I’ve Built with Them.

I run a six-figure Substack business. I’m one person. I don’t have a VA. I don’t have a copywriter. I don’t have a developer on retainer. I don’t have a social media manager, a graphic designer, or a podcast editor.

I have a laptop, a set of AI tools, and — I mean this seriously — two AI thinking partners that know my voice, my brand, my audience, and the way I talk in real life better than any contractor I’ve ever hired.

Here’s everything. The full stack. No gatekeeping, no “DM me for my setup.” Just the tools, the costs, and what each one does in my business every week.

The Two That Run Everything

Before I get into the full list, I need to be clear about something: two tools carry 80% of my business. Everything else is important, but these two are the foundation andt he powerlifters of the group.

Perplexity Pro — $20/month

Perplexity is where I research, plan, and think. It searches the live web, cites its sources, and gives me current, verifiable information — not training data from two years ago.

But the real power is in what Anthropic calls “Computer” — an AI agent inside Perplexity that handles multi-step projects autonomously. I’ve used it to:

• Run a complete SEO audit of my entire website — every page, cross-referenced against my Google Search Console data, with a prioritized fix list split into “needs a developer” and “I can do this myself”

• Rewrite the SEO fields on my top-performing Substack articles — all six fields across five articles, that’s 30 pieces of optimized copy

• Build my entire SEO, AEO, and GEO strategy from scratch — a master document covering traditional search optimization, answer engine optimization (showing up in AI-generated answers), and generative engine optimization

• Research competitors, find what’s ranking, and identify content gaps I should fill

This isn’t a chatbot. This is a research partner that can take a complex project and work through it step by step while I do something else. The work Perplexity has done for me would have cost thousands if I’d hired an SEO consultant.

Claude Pro — $20/month

Claude is where I write. Everything.

Newsletter drafts. Email sequences. Sales page copy. Landing page content. Podcast show notes. Social captions. Offer descriptions. Course outlines. Every piece of written content in my business either starts in Claude or passes through it.

The reason it works so well — and the reason I stopped using every other AI writing tool — is Claude Projects. I have a project set up for 9-to-Thrive with my brand voice, my tone guidelines, my specific rules (don’t use certain words, don’t sound corporate, match the way I actually talk). Claude carries that context across every conversation. I never re-explain myself.

I also use Claude Skills — custom instruction sets that handle specific types of tasks. I have skills for:

• Every type of newsletter I write

• Every type of image I create

• SEO optimization workflows

• Email sequence frameworks

• Content generation patterns

These are reusable, consistent, and they get better over time as I refine them. They’re one of the biggest unlocks in my workflow.

And there’s Claude Cowork — the autonomous task mode in the desktop app. Instead of a conversation, I hand it a project and it runs with it. Compiling research, building documents, reorganizing content. It’s included in the Pro subscription.

The Thinking Partner Angle — This Is the Part That Matters

I could list tools and tasks all day. The thing that actually changed my business isn’t any individual feature. It’s that Perplexity and Claude function as thinking partners.

When I was building my 2026 business strategy in December, I didn’t just sit alone with a notebook. I talked it through — revenue goals, offer structure, content plan, growth targets — and got real pushback, real questions, and real clarity.

When I needed to decide whether to restructure an existing paywalled article or write a completely separate free version for SEO, I didn’t just guess. I worked through the trade-offs, looked at what was ranking, and figured out the right move.

When I’m building a new offer, pricing a new product, scripting a podcast episode, or planning a launch — I’m not doing it alone. I have two partners who know my business, my audience, and my voice, and they’re available at 11pm on a Sunday when I’m in the middle of a sprint.

I’ve hired coaches. I’ve paid for masterminds. I’ve worked with strategists. None of them knew my business as well as these two tools do right now. And none of them cost $20/month.

The Full Tech Stack

Here’s everything else, with real prices.

Wispr Flow Pro — $15/month

This one has taken me to the next level. Wispr Flow is a voice dictation tool, and it’s done something I didn’t expect: it’s trained my AI tools to write the way I speak.

Here’s how that works: I talk naturally — the way I’d explain something to a friend — and Wispr captures it. That raw, spoken version of my ideas becomes the input for Claude. The result is content that sounds like me in a way typed prompts never quite captured. My newsletters, my course content, my sales copy — it all got more natural, more conversational, and more authentically mine the moment I started using Wispr Flow as my input layer.

If you’ve ever read AI-generated content that technically says the right things but sounds like nobody in particular? Wispr is how I fixed that.

Ideogram Plus — $20/month

Every thumbnail and header image for 9-to-Thrive runs through Ideogram. Unlike most image generators, it handles text inside images — which is critical for newsletter thumbnails where the headline is part of the visual.

My thumbnails are bold, colorful, and designed to stop a scroll: hot pink backgrounds, chunky white type, illustrated elements like price tags and gold coins. Ideogram gets me there consistently.

Google AI Pro — $19.99/month

This gives me access to Nano Banana for infographics, ads, promotions, composites and other random image generation. I use it alongside Ideogram depending on the style I need — different tools have different strengths, and having both means I’m never forcing one tool to do something it’s not great at.

Gamma AI — $10–20/month

For presentations, slide decks, PDFs, and even quick prototype websites. When I need a polished visual document and don’t want to spend hours in PowerPoint or hire a designer and wait weeks. Gamma handles it in minutes.

Lovable — $25/month

This is where things get interesting. Lovable is an AI app builder, and I use it to vibe code my own tools. Real tools. Tools I sell and tools I use daily in my business. More on that in a minute.

WriteStack — ~$24–33/month

I use WriteStack for scheduling and managing my Substack Notes. Notes are responsible for a huge percentage of subscriber growth, and WriteStack keeps me consistent without manually posting every day.

Canva Free — $0

Still use the free plan for quick graphics, resizing, and things that don’t need AI image generation. It’s a workhorse.

What I’ve Built

This is the part that gets me excited, because this is what changes the math for solo creators.

I don’t just use AI tools. I build them. Using Claude, Perplexity, Lovable, and my own skills and prompts, I’ve created:

Revenue-generating tools:

• The Creator Profit Engine — a complete brand building studio at thecreatorprofitengine.com. I built this. It creates your brand messaging, ICP, does market research, finds hidden market opportunities, builds your YouTube video strategy (including scripts), provides digital products ideas and creates mid-ticket and high-ticket offers and launch creative.

• The Newsletter Pricing Calculator — helps creators figure out what to charge for their newsletter

• The Offer Stack Builder — helps creators structure their paid offerings

• A Landing Page Content Generator

• A Brand Launch Kit Generator

• A YouTube Video Script Creator

Internal business tools (Claude Skills and workflows):

• An email writer skill that drafts sequences in my voice

• Image generation skills for both Claude and Perplexity

• Skills for every type of newsletter I publish — How-To Guides, paid and free content in different formats

• Skills for every type of image I create — text-led cards, lifestyle photos, hand-sketch concepts, brand promos

Two years ago, building any one of these would have required hiring a developer. Now I build them myself, and I sell them to my audience. The tools that run my business are also products in my business.

The Real Numbers

Let me add it up:

Under $175/month for the entire tech stack that runs a six-figure solo business.

Now think about what this replaces:

• A VA for research, audits, scheduling, and admin — conservatively $1,500–$2,000/month

• A copywriter for newsletters, emails, sales pages, and SEO copy — $1,500–$3,000/month

• A developer for building tools and products — $3,000–$5,000/month for even part-time work

• An SEO consultant — $500–$2,000/month

• A graphic designer for thumbnails and visuals — $500–$1,000/month

That’s $7,000–$13,000/month in services that $175/month now covers.

The subscription cost isn’t even the point. The point is that a solo creator in 2026 can operate at a level that used to require a team of five.

What I’d Tell You If You’re Starting Out

There has never been a better time to run a business alone.

Not because you should do everything yourself forever — maybe you’ll hire eventually. But because the gap between “solo creator figuring things out” and “established business with real systems” has never been smaller.

Start with two tools: Perplexity for research and thinking, Claude for writing and building. Learn to use Projects and Skills in Claude so your context carries over. Add Wispr Flow early — training your AI on your real voice is a multiplier that compounds over time.

Then build. Don’t just consume AI tools — build with them. The calculator that helps your audience? Build it. The email sequence framework you keep reusing? Turn it into a skill. The brand guide that makes your content consistent? Systematize it.

The creators who will do well in the next two years aren’t the ones who found the perfect tool. They’re the ones who learned to build their own.

If you’re ready to stop tinkering and actually build your Substack into a business . . .

the Substack Setup Sprint is where I’d start. It’s four done-with-you sessions where we get your publication, welcome sequence, homepage, and paid tiers set up the right way — so element works towards conversions instead of sitting there half-finished.

And if you’re not sure what your next move should be, take the Start Here quiz. It’ll point you to the right resource for wherever you are right now — whether that’s your first 100 subscribers or your first $1,000 month.

It was exhausting enough that I’d avoid it, which is a spectacular problem when your business runs entirely on content.

I ended up hiring it out to a content team for $4,000 a month.

Now I do it in under 60 minutes. Same quality. More output. Two tools.

Free subscribers get one quick win every week. Want the full deep-dive strategy, monthly coaching, and the Creator Tool Vault? Upgrade to paid →

Subscribe now

Here’s the production line.

Step 1: Research with Perplexity.

Before I write a word, I mine my emails, DMs, and comments to find what my readers are struggling with and pick a topic. Then I ask Perplexity to surface the 5 biggest frustrations my audience has around the topic. Not what I think they want to know. It tells me what they’re actually typing into search at 11 pm when they’re annoyed.

Perplexity pulls from live sources. Real objections, unexpected angles, questions I hadn’t thought to ask. I keep what feels most helpful for my audience, and delete the rest.

Five minutes. Done.

Step 2: Draft with Claude Sonnet 4.6.

I take Perplexity’s research, add my outline and voice-to-text observations and stories, and hand it to Claude. I’ve trained it well so it knows my voice rules, my banned words list, and a specific prompt for the format I request. Claude produces a fast first draft.

Here’s where people go wrong: the draft is a starting point. Not a finished post.

I go in, rewrite the opening completely, add my story, throw in the opinion Claude won’t say without being asked, and cut anything that reads like a LinkedIn post from 2019.

AI = speed. My voice + expertise = the reason people stay subscribed.

Step 3: One post becomes many.

After publishing, I hand the piece back to Claude and ask for three Substack Notes, five hook variations, and a LinkedIn post. What took me 6 hours now produces a week of content in about 30-45 minutes.

And if a post performs? I expand it.

Three of my highest-performing newsletters are now digital products sitting on Gumroad, earning money while I do nothing. That’s the part nobody’s talking about yet. Your best content is already a product — it just needs 90 more minutes and a checkout page.

This week’s experiment:

Take your last newsletter post. Paste it into Claude and type: “Turn this into 3 Substack Notes hooks and one LinkedIn post in [describe your voice].” Watch how much time you just got back.

If your best content feels like it should be a digital product, but you’re not sure how to flip it, my Digital Product in a Day Workshop is March 10th, 10am MT.

You’ll walk away with a ready-to-build digital product. The question is whether you want to figure that out alone or build it live with me in 90 minutes.

Limited spots available.

RESERVE YOUR SPOT NOW

—Carrie

P.S. AI is the production line. YOU are the product. Don’t let anyone — including the algorithm — convince you those are the same thing.

Free subscribers get one quick win every week. Want the full deep-dive strategy, monthly coaching, and the Creator Tool Vault? Upgrade to paid →

Subscribe now

Other ways I can help

→ Substack Newsletter Audit — I will review your Substack homepage and provide actionable feedback on messaging, positioning, monetization strategy, conversion potential, and provide feedback in a video that is yours to keep. plus you’ll get a 90-day implementation plan

→ $10K Offer Prompt Pack — The AI prompts built specifically for creators who know their content is worth more.

→ The Substack Growth Hacks Playbook — What’s working right now, not six months ago.

→ The Creator Cash Flow Club — Access to Substack 360, the #1 system for building a scalable business with multiple income streams, my Creator Tool Vault, all my paid content and master classes, a monthly coaching call + priority DM support.

Good Reads

Before you go — a few things worth your time:

Nine months ago, I paid a designer $500 to create a lead magnet.

It was a beautiful PDF. Gorgeous colors, perfect fonts, professionally branded.

And it converted at 6%.

Last month, I built an interactive calculator using an AI tool. Took me 10 minutes and cost me $0.

It’s converting at 27%.

Not because the design is better (it’s not). Not because I’m some coding genius (I can barely spell JavaScript).

It’s converting because it solves ONE specific problem for ONE specific person in real-time.

Static PDFs are dead. Interactive tools are what people actually want.

And here’s what nobody’s telling you: You don’t need a developer to build them anymore.

Why PDFs Don’t Work Anymore

If you’re still creating PDFs as lead magnets, you’re competing with 10 million other creators doing the exact same thing.

“7 Tips to Grow Your Newsletter”
“The Ultimate Guide to [Whatever]”
“5 Steps to [Generic Result]”

These get downloaded once, opened never, and forgotten immediately.

Meanwhile, interactive tools—calculators, quizzes, assessments—get:

• Used multiple times by the same person

• Shared with friends (”You HAVE to try this”)

• Bookmarked for later reference

• Remembered when it’s time to buy

They position you as the expert who built something helpful, not just another person with a Canva template.

The problem is that many people believe that building these requires hiring a developer for $2,000+ or learning to code.

That was true 12 months ago. It’s not true anymore. Let me show you how I do it.

If you’re ready to grow your business by building multiple income streams from one newsletter, you’re in luck because I’m offering a 35% off an annual membership through December 1.

Get 35% off forever

What I Wish I’d Known 9 Months Ago

Here’s what took me from spending hours (and dollars) on lead magnets to creating them in less than 30 minutes:

Read more