Who you are, who you write for, the format you like, the words you’d never be caught using. The block you’ve typed so many times you keep it in a notes file just to copy it faster.

Then you ask for the thing. And half the time the draft still comes back sounding like a LinkedIn post wrote it.

So you’re re-training it. Every session. From scratch.

Here’s the good news: you don’t have to anymore. Two of the biggest AI tools just made the same move within months of each other. You can teach the thing a task one time and it keeps it. FOREVER. No more re-paste.

By the end of this you’ll know the difference between a prompt and a skill, the two ways to build one, and the exact task you should turn into your first skill tonight.

The tax you didn’t even know you were paying

Every session starts at zero.

The AI doesn’t remember that you write for solo creators and not enterprise marketing teams. It forgot you hate em dashes. It lost your format, your structure, the exact phrasing you’d never use. So you tell it. Again. Like you didn’t say the same thing yesterday.

Call it the re-explain tax. 5 minutes here, 10 minutes there, every time you sit down. If your writing window is the 90 minutes before the kids wake up, you just burned a chunk of it dragging the AI back to where it was last Tuesday.

And doing it by hand is sloppy. You forget a rule. You paste a slightly older version of your style notes. The output drifts. One day it sounds like you. The next day it sounds like everyone.

That’s the real cost. Not the minutes. The drift.

What a skill actually is

A prompt is a paper towel. You use it once, you get your output, you throw it out. Tomorrow you grab a fresh one and start over.

A skill is the thing you build once and keep.

It’s a saved set of instructions the AI loads on its own when the matching task shows up. You don’t paste it. You don’t remind it. It sits there, and it pulls itself in when the job calls for it.

Anthropic shipped this for Claude last fall. A skill is a folder with a file inside that spells out how to do a specific task the way you want it done. Claude reads the name and short description of every skill you’ve got at the start of each session, so it knows what’s on the shelf, then loads the full instructions only when it needs them. By December they’d turned it into an open standard, which means a skill you write for one tool can run in others too.

Plain version: you write your rules down ONCE, in a way the machine reads on its own, and you stop being the one who has to remember them.

Two ways to teach it once

Pick based on whether the task is easier to describe or easier to show.

Write it. If you can explain the task in words, write the skill. Open a doc. Spell out when it applies, what it needs from you, the steps to follow, and how to tell if the output is actually right. That’s a skill. Your whole style guide, the one you keep re-pasting, becomes a skill that loads itself every time you draft.

Show it. This is the new part. On June 18, OpenAI added Record and Replay to its Codex app. You do the task once on your Mac while it watches. Then it writes the whole thing out as an editable skill you can open and fix later: when to use it, what it needs, the steps, how to check the result. You can run it again with new inputs whenever the task comes back.

The “show it” path is built for the fiddly stuff that’s a pain to write out. Pulling the same report every Monday. Or publishing a piece through the same clicks you always do, in the same order, the kind of thing that’s faster to do than to describe.

Two honest catches on Record and Replay. It’s macOS only right now, and it isn’t available in the UK, the EEA, or Switzerland yet. If that’s you, the write-it path works fine and gets you most of the way there.

What actually goes inside (and why your voice stops drifting)

Every good skill answers 4 questions. When do you use this. What does it need from you. What are the steps. How do you know it worked.

That last one carries more weight than people give it. The “how do you know it worked” line is the check that keeps the output honest. Skip it and you’ve automated a task with no way to catch it when it goes sideways.

Here’s where this lands for anyone who writes for a living.

Your biggest fear with AI is that it flattens your voice. Sands you down until you sound like the feed. And that happens because you’re re-explaining your voice from memory every session, badly, and the AI fills the gaps with generic mush.

Put your voice rules in a skill and that stops. The banned words. The rhythm. The structure. The phrasing you’d never let through. Written once, loaded every time. The draft starts from your standards instead of the internet’s average.

[STORY HOOK: Right here is where one line about YOUR own voice skill would land hard. Something like the first time you ran a draft against your voice rules as a skill instead of pasting them in, and the output finally held without you babysitting it. Two sentences, a real beat. It proves the “voice stops drifting” claim with your own receipts instead of theory. Fallback: I keep this section in second person and it still works.]

Don’t turn your thinking into a skill

This is where people are going to overdo it.

A skill is for the task you do the SAME way every time. The repeatable stuff. Formatting, structure, your style rules, the boring report you pull every week. Anything with a fixed path and a right answer.

Judgment is a different animal. Picking what to write about this week is not a skill. Reading a reply from a subscriber and knowing what they actually meant is not a skill either.

The line is simple. If the task changes every time and needs you to think, keep doing it yourself. If the task is the same every time and you’re just the one remembering the steps, build the skill.

Skill the grunt work. Keep the thinking.

Start with the thing you’ve typed 10 times

You already know your first skill.

It’s the paragraph you paste in at the start of every session. The instructions you’ve typed so many times you could say them in your sleep.

Take that. Put it in a skill. Never paste it again.

Then watch what else you repeat. Every time you catch yourself explaining the same thing to the AI twice, that’s a candidate. Build it as you hit it. Soon the AI stops making you start from zero.

You open Claude. And this time you don’t paste anything. You just ask for the thing.

And it already knows.

Ryan

P.S. If you want the actual skills I run behind this newsletter, the voice rules and the structure doing the quiet work, that’s what I hand paid subscribers. Same idea as the article: build it once, use it forever.

TL;DR: Google made Gemini 3.5 Flash the default model behind Google AI Mode in May 2026 and pushed it to nearly 200 countries. For writers, it turns Search into a fast research engine that fans a single question into a dozen-plus parallel searches and answers with citations. The speed is genuinely useful for research and fact-checking. The catch: it scored a 61% hallucination rate on one accuracy benchmark, so treat every answer as a lead to verify, not a source to trust. Below: how it works, where it breaks, the Deep Search and Personal Intelligence power moves, and how to get your own writing cited inside AI answers.

If you write for a living and you still picture Google as a list of blue links, your mental map is 6 months stale.

In May, Google made Gemini 3.5 Flash the default brain behind AI Mode. Then rolled it out to nearly 200 countries.

Most writers I know haven’t touched it. They’re still pasting research questions into a chat window in another tab.

This piece tells you three things. What AI Mode actually is now. Whether it’s good enough to pull into your writing day. And where it quietly falls on its face.

Short verdict up front: worth opening for research. NOT worth trusting blind.

The speed is real. The accuracy is shakier than Google wants you to believe.

What Google AI Mode actually is now

AI Mode is the tab inside Google Search that stopped handing you links and started handing you answers.

You type a question. Gemini reads across a pile of sources. It writes you back a synthesized answer with citations you can click. Ask a follow-up and it remembers what you just asked.

Google flipped the default to Gemini 3.5 Flash in May at I/O 2026. The number they keep repeating: AI Mode crossed 1 billion monthly users, with queries more than doubling every quarter (blog.google).

The part that matters for you is under the hood.

When you ask a question, the model doesn’t run one search. It uses query fan-out, splitting your question into a dozen or more smaller searches that all run at once, then stitching the answer together (Search Atlas).

So you ask “how have newsletter open rates shifted since Apple Mail Privacy Protection.” It isn’t reading one blog post and paraphrasing. It’s fanning out across a dozen sources and merging them.

For the messy, multi-part questions writers actually have, that’s useful.

Why Gemini 3.5 Flash’s speed changes how you research

Gemini 3.5 Flash is built to be fast. Google says it pushes out tokens about 4 times faster than other frontier models (blog.google).

You feel it. Ask a real research question and the answer lands in seconds. No spinner. No alt-tab to go check email while it thinks.

Speed changes behavior.

When research is slow, you batch it and cut corners. When it’s instant, you actually check the thing you were about to assert. You ask the follow-up. You chase the second source.

The friction that used to talk you out of the work just isn’t there.

That’s the real pitch. Not that it’s smart. That it’s fast enough to stay in your flow.

The Gemini 3.5 Flash cost reality nobody mentions

For you, typing questions into Search, AI Mode is free. No subscription. 98 languages, almost 200 countries (blog.google).

But there’s a number worth knowing.

Google tripled the token price. Gemini 3.5 Flash now runs $1.50 per million input tokens and $9.00 per million output, triple what the old Gemini 3 Flash charged at $0.50 and $3.00 (The Decoder).

Why care if you’re not a developer?

Because “free” search products tend to get monetized. Universal Cart already wires shopping straight into AI Mode across Search, Gemini, YouTube, and Gmail, sitting on a catalog of more than 60 billion product listings (blog.google).

The tool is free because you’re being walked toward a checkout. Use it knowing that.

Where Gemini 3.5 Flash falls on its face on accuracy

This is the part the launch posts skip.

Gemini 3.5 Flash scored a 61% hallucination rate on one Artificial Analysis benchmark. That’s an improvement, down 31 points from the older Flash model. It still trails the leaders, MiMo-V2.5-Pro and Grok 4.3, both sitting around 25% (Artificial Analysis).

A 61% rate on that test means it still invents something on most of those questions. Sit with that before you trust an unsourced claim it hands you.

It also isn’t a clean upgrade. The Flash model still loses to the older Gemini 3.1 Pro on harder reasoning tests like Humanity’s Last Exam and the long-context MRCR v2 benchmark (Appwrite).

Google made the faster model the default. Not the smartest one.

For quick lookups, fine. For a 12-source piece where one wrong fact tanks your credibility, slow down and verify.

My rule: treat every answer as a lead, not a source. Click the citation. No citation, assume it might be inventing the detail.

That one habit separates writers who use AI Mode well from the ones who publish a confident lie because a search box told them so.

Want the part that actually changes your week?

Everything above is the honest review. What’s good, where it breaks, what to watch for.

What’s below is how I actually USE it.

The power-user moves most people never find. Plus the AI Mode features that change how your own writing gets discovered now that Google answers questions instead of listing links.

Paid subscribers get the rest:

• The Deep Search trick that turns AI Mode into a research analyst running hundreds of searches and handing you a cited report

• How to wire Gmail into AI Mode so it answers using your own receipts, bookings, and threads

• The content structure changes that get your writing cited inside AI answers, backed by AirOps’ 2026 analysis

• The one freshness number that decides whether AI Mode ever quotes you

• My verify-or-kill checklist for using AI Mode research without publishing a hallucination

[Upgrade to read the rest →]

Subscribe now

Read more

Subscribe now

The systems make sense. The logic clicks.

But you haven’t built one yet.

That gap is what paid solves. Free shows you the system. Paid hands you the build, tested and working, ready to drop into your business by Tuesday afternoon.

Here’s what shows up the day you upgrade.

Subscribe now

The Expert Profile Library

Drop-in profiles you paste into any prompt so the AI writes like David Ogilvy, Gary Halbert, Robert Cialdini, or any of the other experts I’ve built. Copy that sounds like the person who taught marketing in the first place. → Browse the library

The full prompt and workflow catalog

Every system I run inside my own business Monday through Friday, copy-paste ready. Indexed by use case (sales, content, operations, customer support). Sunday night you realize you forgot to write Tuesday’s newsletter. You pull “newsletter draft” off the shelf. 8 minutes later you’ve got something that sounds like you.

Every back issue, searchable forever

Free subscribers see what hit their inbox that week. Paid sees the entire archive. Every prompt and workflow I’ve ever published, searchable from the first issue. Problem shows up at 9pm Thursday, you pull the answer instead of waiting for me to write about it again.

Direct reply access

I read every email. Send me your stuck workflow and you’ll usually get a fix back the same week. This is the closest thing to consulting you’ll get for $15 a month.

When the problem outgrows what email can solve, deploi.co is where it goes. Most readers never need that. Reply access handles almost everything.

The math

One system that returns 2 hours a week returns 100+ hours a year. $15 a month or $100 a year. Even if 4 out of 5 systems don’t apply to your business, the one that does pays for the subscription 50 times over.

Subscribe now

Ryan

Your feed is on fire again.

Another model dropped. The thumbnail brigade is back with “THIS CHANGES EVERYTHING” and benchmark screenshots you can’t read and a YouTube hype boy telling you how shocked he is from this new release.

Like this guy:

No Alex, it’s not…

I’m not going to do that to you.

If you use Claude for actual work and you just want to know whether this release matters to you, I am going to drop it for you, in plain English.

No benchmarks. No charts. Just what changed and whether you should care.

Here’s the verdict up front: Claude Opus 4.8 is NOT a leap.

Anthropic said so themselves. They called it a “modest but tangible improvement.” When the company that built it is telling you to lower your expectations, believe them.

But “modest” doesn’t mean “useless.”

A few changes fix things you’ve been annoyed by, probably without realizing they were fixable. Those are worth your time.

Let me walk through them.

Claude 4.8 hallucinates less

The internet if full of tales where people got caught submitting fabricated AI hallucinations in their work.

Like the lawyers who were sanctioned for filing legal briefs full of fake AI citations.

You ask Claude something. It hands you a clean, confident answer. Then you go check, and half of it was invented. No warning. No “I’m not sure here.” Just smooth, total fiction delivered like fact, almost as smooth as my 10 year old trying to convince me she wasn’t the one who didn’t put her dishes away.

Anthropic put most of their effort this release into that exact problem. They trained 4.8 to flag when it’s unsure and to stop making claims it can’t back up. Early testers say it owns its uncertainty more and fabricates less.

Is it better? I don’t know yet. Nobody does. I have been trying to break it, but I haven’t come across a) hallucinations or b) it telling me it’s unsure.

A company telling you their model is more honest means nothing until you’ve watched it admit “I’m not certain about this” on a question it would’ve bluffed last month. So test it.

The single biggest reason to distrust AI is that it lies to your face with a straight one. hopefully this fixes that.

The new effort dial: you control how hard Claude thinks

Ever ask Claude to do something dead simple, like reformat a list, and watch it sit there reasoning like you’d asked it to explain the meaning of life?

That was a real complaint about Claude 4.7. It would burn time and tokens deliberating over tasks that needed none of it. Annoying when you’re in a hurry. Worse when you’re paying per token.

Claude 4.8 adds an effort dial.

It lives right next to the model picker in the Claude app and in Cowork. Think of it as how hard you want Claude to think before it answers.

Crank it up when the problem is genuinely hard and you want it to chew on the thing. Drop it down when the task is obvious and you just want the answer.

Note that it’s set at a “high” default. If you are using Claude for simple tasks, you may want to lower this to lower costs.

The bigger context window: long documents finally stay in memory

Claude 4.8 can now hold a massive amount of text at once. A full year of meeting notes. A 300-page report. An entire manuscript or your whole newsletter archive.

All in one go.

The kind of thing where you used to paste it in and watch Claude forget the beginning by the time it reached the end.

If you work with long documents, this is the change you’ll feel. Drop the whole thing in. Ask your questions. It remembers what was on page 4 when you’re asking about page 280.

Fast mode: pay more, wait less

Small one, but worth knowing.

There’s now a fast mode that cranks the speed up about 2.5x. You pay a premium for it. Pay more, wait less.

Most people won’t need it. But if you’re ever watching the cursor blink while a deadline breathes down your neck, the option’s there.

Dynamic workflows (skip this section if you don’t code)

Anthropic also shipped a thing called dynamic workflows. It lets Claude run a huge number of helper agents at once on giant coding jobs inside Claude Code.

If that sentence meant nothing to you, good. It doesn’t apply to you.

Move on with a clear conscience.

So should you care about Claude Opus 4.8?

A little.

Switch to Claude 4.8. It’s free, same price as Claude 4.7, and it’s better at the stuff that frustrates you most.

Play with the effort dial. It’s the one new toggle you’ll use.

And if you ever work with big documents, this is the upgrade that quietly makes your week easier.

That’s the honest read. Not a revolution. Just a model that lies less, wastes less of your time, and forgets less of what you give it.

Which, when you’ve spent the last year fighting all three of those, is more than enough reason to flip the switch.

Common Questions About Claude Opus 4.8

Is Claude Opus 4.8 worth switching to from Claude 4.7? Yes, for most people. It’s the same price, available in the same places, and the changes (less hallucination, the effort dial, the bigger context window) hit the parts of using Claude that annoy you most. There’s no reason to stay on 4.7 unless you’ve built something that depends on its specific behavior.

Does Claude 4.8 cost more than Claude 4.7? No. Same pricing tier. The only thing that costs extra is the new fast mode, which is optional and runs about 2.5x faster for a premium.

What is the effort dial in Claude 4.8? A new toggle next to the model picker in the Claude app and in Cowork. It controls how hard Claude thinks before answering. High effort for complex problems. Low effort for simple tasks where you just want the answer. The default setting reports are mixed, so open your own dropdown and check.

Does Claude 4.8 still hallucinate? Yes. Every large language model still hallucinates. Anthropic trained 4.8 to flag uncertainty and fabricate less, and early testers say the change is real. But you should still verify anything that matters before you use it. The improvement is that 4.8 is more likely to admit when it doesn’t know.

How big is Claude 4.8’s context window? Big enough to hold a 300-page report, a year of meeting notes, or a full manuscript in a single prompt without losing the beginning. If you work with long documents, this is the upgrade you’ll feel first.

What are dynamic workflows in Claude 4.8 and do I need them? Dynamic workflows let Claude run many helper agents in parallel inside Claude Code on large coding jobs. If you don’t write code, you can ignore this entirely. It doesn’t show up in the regular Claude app or in Cowork for non-coding work.

Where can I use Claude Opus 4.8? The same places you used 4.7: the Claude app, Claude Code, and Cowork. Switch the model in the picker and you’re on it.

About the author

Ryan Stax writes The AI Handbook, a Substack newsletter for solo content creators and newsletter writers who want to use AI in their work instead of just reading about it. He’s the founder of deploi.co and has been building with Claude since its early releases.

Related reading: Claude Managed Agents: What You Need to Know, the honest breakdown of the last Claude release worth reading. Also worth a read: How to Use Claude Cowork Better Than 99% of People, the complete 10-minute setup for the surface where the new effort dial lives.

If you are still using it like a glorified Google, you are missing the point (and likely wasting a lot of money).

You may as well hire a chief of staff to man your reception desk.

Below: the best Claude features as of May 2026, tagged for whether an online entrepreneur should adopt it tonight, bookmark it for the next launch, skip it until you hire a team, or write it off as YouTube hype.

Some of these may be new to you, you may know all of them. Either way it’s good to refresh your list, bookmark these and reference them later.

17 features, 2 or 3 sentences each. 5 minutes start to finish.

1. Memory import

If you’re moving from ChatGPT, Claude lets you suck your memory across in about 2 minutes. Settings, Capabilities, Start Import. Worth doing once if you’ve been talking to ChatGPT for years and want Claude to know what it already knows.

2. Model selector

Bottom right of the chat window. Haiku for fast and cheap, Sonnet for almost everything, Opus when you need it to think hard, or adaptive mode if you can’t be bothered deciding. If you’re drafting a launch strategy, use Opus. If you’re cleaning up subscriber export data, Haiku is fine.

3. Gmail connector

Sidebar, Connectors, Gmail. Claude can read, search, summarize, and draft replies inside your inbox. For an online entrepreneur this is mostly useful for triaging customer support and clearing the pitch pile without opening Gmail at all.

4. Calendar connector

Same path, Google Calendar instead. Useful if you batch your work blocks and want Claude to find time for you. Honestly less load-bearing than the email one if your calendar is mostly your own.

5. Artifacts

Tell Claude to “build me an editable spreadsheet of my 12 most-recent sales calls with columns for objection, response, outcome, and follow-up” and it makes a working mini app right in the chat. Great for pipeline tracking, offer audits, customer experiments, and seeing which hooks actually convert. The thing builds the thing while you describe what you want.

6. Interactive visuals

Say “visualize this” and Claude renders a clickable diagram of whatever you just gave it. Useful when you want to map out a customer journey or sketch how an offer funnel actually flows without opening Figma. The visual is throwaway, the thinking is the point.

7. Projects

THIS is the one. Sidebar, Projects, New Project. Drop in your offer docs, your brand voice guide, your customer interview transcripts, your last launch debrief, and now every chat in that project has all of it as context. One project per offer or per ongoing client. STOP pasting the same 4 paragraphs of context into every new chat.

8. Voice mode

Waveform icon on desktop, mic on mobile. Talk through your idea while you walk, get the full transcript at the end. Useful when you can talk faster than you can type, which is most of the time.

9. Chrome extension

Claude for Chrome from the Web Store. The extension can read pages you’re on, scrape competitor pages, pull data from ad libraries, and do research without you babysitting it. Online entrepreneurs can use it for fast competitor teardowns and pricing research without copy-pasting.

10. Cowork

Desktop app, top left, toggle computer use on first. Claude takes over your machine and runs multi-step tasks while you walk away. Set it up to draft your weekly broadcast, queue 5 social posts, pull your funnel numbers into a doc, and send the summary to your team Slack while you go make coffee.

11. Scheduled Task

Inside Cowork, type /schedule. Runs the same task every day as long as your computer is on. The best use for an online entrepreneur is a morning ops brief: scan industry news, pull what’s hot in your niche, summarize last night’s customer messages, and queue your top 3 priorities in your inbox before you sit down.

12. Claude Dispatch

Mobile app paired to your desktop Cowork via QR code. Useful if you spend time away from your desk and need to trigger something running on your home machine. Honestly skippable for most online entrepreneurs unless you travel a lot or operate from cafes.

13. Claude Code

Claude as a software developer that builds tools in plain English. Most online entrepreneurs don’t need this until they need it, which usually looks like wanting a custom CRM view, a lead scoring script, a tool that scrapes pricing changes, or a private mini app for tracking offers. Bookmark, return when you have a specific itch.

14. Claude Channels

Use iMessage, Telegram, Discord, or another messenger to text Claude Code from anywhere. Niche unless you’re already deep in Claude Code. Online entrepreneurs can mostly skip this until you want to trigger ops from your phone at 11pm.

15. Claude Skills

Customize, Skills, Browse. Skills are repeatable instructions Claude follows the same way every time. Build one for your brand voice, one for your sales page rewrites, one for turning customer calls into case studies, one for cleaning up survey responses into a tagged spreadsheet. If you only adopt one feature from this entire list, adopt this one.

16. Claude Design

claude.ai/design. Builds pitch decks, one-pagers, landing page mocks, motion graphics. For an online entrepreneur the best use is sales page mocks, lead magnet covers, pitch deck drafts, and ad creative variations in your brand style, fast.

17. Custom instructions and Style

This is the one Martell MISSED. Inside Settings, you can give Claude permanent custom instructions and feed it writing samples to train on your voice. Drop in 10 of your best emails and sales pages, set the rules for how you want Claude to respond, and every chat starts with the brand voice already locked in.

That’s the tour. 17 Claude features, filtered through the test of whether they help one online entrepreneur run a tighter business.

Set up Projects and Skills this week.

That puts you ahead of anyone who tried to wire up 14 of these and burned out on connectors they don’t use.

Tell me in the comments which one you actually opened first. I’m curious where the real bottleneck is.

Newsletter, course, agency, info product, coaching, SaaS, whatever the wrapper is.

You do everything inside the business. You’re the salesperson. You’re the bookkeeper. You’re customer support at 11 PM because nobody else will do it.

You’ve likely found that one ZOMBIE task that eats away at your energy every single day. You know the one, it’s the one you dread, the one you put off and it’s the choke point every time you sit down.

You’ve tried tools. You bought the Notion template. You wired up the Zapier. You watched the YouTube walkthrough. Two weeks later you stopped using any of it and went back to doing the thing manually.

These are the tasks that “agents” can take care of. And I’ll be honest they are NOT scary to set up.

By the end of this article you’ll have 7 agents other online entrepreneurs are running right now. You’ll see the pattern that tells you which of your problems is solvable, and the tool that makes building one possible without code.

One note before the list. The Hyperagent link I’ll drop a few times is my referral. Use it and you start with $1,000 in free credits. Plenty of runway to build, break, and rebuild a few agents before you spend a dollar.

Agents aren’t scary.

To be honest, they are actually easier to setup than you may think.

An agent is an employee with infinite shifts. You give it the goal. It figures out the steps. It picks which tools to use and connects itself to the tools you already use. It does the actual work. When you wake up the work is done.

Tell it what you want. It handles the how.

For an online entrepreneur, this is the unlock. You’ve never been able to afford the junior employee who would do the boring weekly job for you. The agent IS that junior employee.

It costs cents per task, never quits and never forgets.

The version of agents you saw six months ago was clunky. n8n. Zapier. Hand-wiring webhooks. JSON schemas you didn’t want to look at. Most non-technical people gave up around step three.

This wave is different. You describe what you want in plain English. The agent figures out the rest.

Which Of Your Problems Is Actually Solvable

You can’t agent your way out of everything YET. Agents work on a specific kind of problem.

Pull out a piece of paper. List the five recurring tasks that eat your week. Then check each one against this:

• It repeats. Daily, weekly, monthly. Same shape every time.

• It has a clear input and a clear output. You can describe both in one sentence.

• It crosses multiple tools. The work touches your inbox, your calendar, your CRM, your spreadsheet, your bank, all in the same job.

• The result is reviewable. You can tell in 30 seconds whether the agent did it right.

The tasks on your list that hit all four are the ones an agent takes. The others wait.

Seven Agents Running In Real Businesses Right Now

Below are seven agents online entrepreneurs are running..

Each one with the same micro structure: the problem, what the agent does, who’s already running it.

Pick the one closest to your worst week.

Subscribe now

1. The Single DM Inbox Agent

You have five inboxes. Email, Slack, LinkedIn DMs, Instagram, X. They check in every twenty minutes. Most of the messages are some 18 year old in India trying to conivnce you he’s the best cold outreach guy.

BUT… the ones that matter get buried under newsletter signups, podcast pitches, and people cold pitching you their garbage service.

The agent pulls every DM across every channel into one feed. Reads each one. Scores it by intent (lead, customer support, fan mail, junk, partner). Drafts replies for the urgent stuff. Sorts the rest into folders. You sit down once a day and clear the whole thing in fifteen minutes.

Who’s running this: creators, coaches, and consultants whose business runs on relationships. Newsletter operators with engaged readers. Anyone whose phone buzzes more than 50 times a day.

2. The Bookkeeping Agent

You hate bookkeeping. You put off your bookkeeping. Last March you spent eleven hours in one weekend panicking through receipts because tax season showed up and you hadn’t logged a single expense since November.

The agent lives in Slack. You type “$60 lunch with prospect Sarah at Earls.” It logs the expense, files the receipt photo, categorizes it, updates the books. Drafts invoices from a one-line prompt. Generates a monthly P&L on the first of every month and drops it in your inbox.

Who’s running this: agency owners, consultants, course creators with multiple revenue streams, anyone running their own books because they can’t justify a bookkeeper yet.

3. The Customer Feedback Synthesis Agent

Your customers are talking. They’re leaving reviews on G2 and Trustpilot. They’re sending support tickets. They’re posting screenshots on X. They’re filling out the survey you sent last quarter. You read maybe 20% of it. The rest disappears.

The agent reads everything. Pulls reviews from every site you sell on. Reads every support ticket. Watches mentions on social. Synthesizes themes into a weekly digest in your inbox. Flags the urgent stuff (churn risk, product bug spike, angry customer about to post on LinkedIn) the moment it shows up.

Who’s running this: SaaS founders, info product sellers, agency owners with retainer clients, anyone with more than 50 customers making noise across more than three channels.

4. The Competitor Research Agent

There are five competitors you should be watching. You’re watching zero of them. The last time you checked their site was March. You don’t know they raised prices in April, launched a new product in May, hired a head of sales last month, or started running ads to the same keywords you’re going after.

The agent watches everything. Tracks pricing changes. Catches new product launches. Reads their content and clips the good stuff. Monitors their hiring pages. Sends a Friday afternoon digest with the moves of the week and what they signal.

Who’s running this: founders in competitive markets, consultants pricing against alternatives, agencies pitching against agencies, anyone whose deal volume gets killed by a competitor’s move they didn’t see coming.

5. The Sales Prospecting Agent

You know who your ideal customer is. You don’t know where to find 200 of them. And if you found them, you don’t have the hours to write 200 personalized emails.

The agent does both. Feed it your ideal customer profile. It hunts on LinkedIn, scrapes the company sites, pulls the right decision maker. Finds the email. Drafts a personalized first-touch in your voice. Sends it. Tracks the open. Drafts a follow-up if there’s no reply in four days. Books the call when they bite.

Who’s running this: B2B founders, consultants, agency owners, anyone whose business needs five new clients next quarter and currently has zero pipeline.

6. The Sponsorship and Partnership Pipeline Agent

You should have sponsors on your newsletter. You should have partners running cross-promos with you. You should have brands paying you to mention them. You have none of it because outreach feels gross and you never get around to it.

The agent does the gross part. Finds brands whose audience overlaps yours. Pulls their contact, their recent campaigns, their budget signals. Drafts a personalized pitch in your voice. Tracks every response. Logs the no’s so you stop pitching them. Flags the maybes for a follow-up in 90 days.

Who’s running this: newsletter operators, podcasters, course creators with engaged audiences, anyone whose business should be running on partnership revenue but isn’t because the outreach work never gets done.

7. The Info Product Development Agent

You have a product you should be selling. You’ve thought about it for months. You know the topic. You know your audience wants it. You don’t have a free month to sit down and build it.

The agent mines your existing material. Reads your newsletter archive, your DMs, your support emails, the comments under your posts. Pulls the questions your audience asks most. Outlines a product that answers them. Drafts the sales page. Builds the launch email sequence. Generates the thank you page. Hands you a near-complete launch you can ship in a week instead of a quarter.

Who’s running this: creators sitting on an engaged audience, coaches with a proven methodology, consultants with a process that’s begging to be productized.

How To Actually Build One

Hyperagent is the platform I’m running all of this on. Built by the Airtable team. Aimed at people who are not engineers.

Four things make it work for online entrepreneurs:

• Triggers and scheduling are built in. You don’t just build a thread. You deploy an agent that runs while you sleep, posts in your Slack, sits in your inbox waiting for the next trigger.

• Apps connect through OAuth. You log in like you would with any other tool. No API keys floating in plain text on your desktop. The security situation that scared you off the earlier agent builders is handled.

• Skills are reusable. Build a “write in my voice” skill once. Every agent you deploy after that uses it. The work compounds the longer you use the platform.

• Performance scoring on every run. Every job gets graded against criteria you set. You see the agent drifting before it starts costing you. This is the difference between a toy and a system you can run a business on.

The $1,000 free credits through the link are enough to build MANY agents and run them for weeks.

What To Do Today

Don’t try to build all seven. You’ll burn out by Tuesday.

Look at your list of recurring tasks. Pick the one that’s costing you the most this week. The one you’re avoiding. The one you’ll be doing at 9 PM tonight if you don’t fix it.

Open Hyperagent. Use the closest agent template. Connect one tool. Run the agent once. See what comes back.

The first version will be rough but keep refining it. Build it anyway. By the second agent you’ll know EXACTLY what you’re doing. By the fourth, you’ll be charging clients to build them.

If you're an overwhelmed online entrepreneur, the paid side is where the receipts live.

The prompts. The agent configs. The templates I'm copying across builds.

Every system I've shipped in the past six months. Plus my DMs are open to every paid subscriber.

Subscribe now

Ryan

The headlines test fine and the topics are solid. Your old posts used to land. Lately the replies are thinner, the shares are gone, and the people who used to email you back are quiet.

Here’s what changed. You started using AI to draft. And your writing started leaking the same fingerprints every other AI-drafted newsletter or X post is leaking right now.

I’m not anti-AI.

I draft with it every week.

The point is to use it as a thinking partner, then strip every detectable trace of it out before you hit publish.

Most of you are doing the first part and skipping the second. Your readers feel it even if they can’t name it.

And it’s eroding trust and credibility.

A 2024 survey by the Reuters Institute for the Study of Journalism found that 52% of respondents feel "uncomfortable" with news produced mostly by AI, citing a lack of transparency and "human touch" as primary drivers for unsubscribing from newsletters

Today I’ll walk you through the obvious tells your audience is likely already picking up on. And I’ve got a skill I’ll hand you to fix them. It’s the same system I run every article through before I ship it, including the one you’re reading right now.

Why Your Brain Goes Numb on AI Writing

Here’s the mechanical reason why AI writing sucks:

A language model writes by picking the most statistically likely next word, over and over, all the way through a sentence. That’s the whole architecture, output is whatever sequence the training data says is most predictable.

Your brain does the same thing when it reads. It predicts the next word a fraction of a second before your eyes get there. When the prediction lands, you process the word with almost no effort. When the prediction misses, your brain has to wake up and actually engage.

AI text lands exactly where your brain predicted it would. Every time. Reading it becomes a skim-coma where you scroll past entire paragraphs without retaining anything because nothing required you to engage.

Your audience is doing this right now with your last six articles. They scroll through without remembering anything because their brains finished each sentence before their eyes got to the period.

High-friction writing breaks the prediction. Every sentence has a turn the reader didn’t see coming. That turn is the only thing keeping the brain awake long enough to remember a single word you wrote.

Now let’s get into what your readers are actually clocking.

The Same-Length Sentence Problem

Read your last article out loud. Listen to the rhythm.

If you can tap a steady beat the whole way through, the article is a metronome. Every sentence comes in at fifteen to twenty words with the same breath, cadence, and shape.

Humans don’t write like that. We sprawl, then we stop short, then we pile on a clause and drop a fragment on the back end.

Three words.

Then sixty.

AI averages every sentence toward the same length because the training data averages out that way on aggregate. Detection tools call this low burstiness. It’s the single strongest statistical tell of AI-written prose, and your readers feel it before they ever consciously notice the pattern.

The Triplet Trap

Count the lists in your last three articles.

Patience, dedication, and resilience. Build, ship, iterate. Plan, execute, repeat. Strategy, mindset, and action. Every list comes in threes because the model thinks three is the magic number.

Two feels incomplete to the model.

Four feels heavy.

Three sits dead center in the comfort zone of the training data, so the model defaults there every time it has to enumerate anything. Your articles end up with triplets stacked across paragraphs like a poker hand.

Once you see this, you can’t unsee it.

The “Not X, It’s Y” Tic

This one is the worst offender. It’s also the easiest one to catch.

“Success isn’t about working hard. It’s about working smart.”

“The question isn’t whether AI will replace us. It’s whether we’ll let it.”

EVERY AI-drafted article on the internet is built on some version of this construction. The model loves it because it sounds profound for free. Editors and trained readers spot it from a mile off. If your article has two of these in the first 300 words, your audience already knows what they’re reading.

Zombie Words

These are the words that signal “professional writing” inside the model’s training data while saying nothing.

Delve. Leverage. Foster. Harness. Tapestry. Landscape. Navigate. Robust. Seamless. Showcase. Underscore. Profound.

You don’t say any of these words at the kitchen table. You probably don’t say them in a Zoom meeting. So why are they in your newsletter? Because the model put them there and you didn’t pull them out.

The Reddit thread on AI writing has a running joke now. Spot the “delve” and you’ve spotted the bot. Your audience is in those threads. They know.

The Floating Concepts

AI floats. It writes about success, freedom, growth, resilience, and transformation, and never once names the cold coffee on your desk at midnight, the Slack message from the client that wrecked your Tuesday, or the Excel formula that broke at 4:17 AM and cost you three hours of sleep.

Floating concepts read like writing. They’re scaffolding wearing writing’s clothes. Real writing has objects, names, dates, and the texture of one specific moment your reader can see in their head. If your last article didn’t have a single proper noun in it, you wrote scaffolding.

The Hedge Habit

Watch how every strong claim gets softened in the next sentence.

“Many experts believe...” “Some have argued...” “A growing number of creators...” “Studies suggest...”

This is the model trying not to offend anyone, because its safety training penalizes confident statements. Your readers came to you for an opinion. When you hedge every claim, you sound like a Wikipedia article that got asked to host a podcast.

Name names, cite real numbers, or own the claim yourself. “I think this” beats “some have argued this” every single time.

The “Nobody Tells You” Setup

These are the contrarian hooks AI loves because they sound like insider knowledge.

“Here’s what nobody tells you about consistency.”

“The thing everybody gets wrong about audience building.”

Both of these promise forbidden knowledge and then deliver something the reader has already seen fifteen times this month. The opening sounds bold. The next paragraph collapses into generic advice. The model has no insider position to speak from, so it borrows the rhetorical move and fills the rest with whatever sat highest in the training data.

If you find one of these in your draft, either replace it with a specific claim you can defend by name and number, or cut it.

The Em Dash Problem

Em dashes are the AI safety net. The model uses them to hedge a claim inside a single sentence without ever committing to a period.

Watch the pattern. “The key to productivity—something most people overlook—is that motivation follows action.”

Two dashes in one line, both used to qualify the claim instead of stand behind it.

Here’s the wrinkle on this tell. Em dashes are a real punctuation mark and some writers use them well as part of their actual voice. If that’s you, you’ll have to tune the skill yourself so it allows them in your work without flagging. The default skill bans them outright because the statistical signal is now too strong on the reader side. Most subscribers in 2026 read an em dash and assume Claude wrote it, regardless of who actually typed the sentence. That’s the cost of using a punctuation mark the model also loves.

The rest of this article is a skill that you copy and paste.

That is the whole product. There is no course you have to sit through. There is no PDF workbook with stock photos of a guy in a blazer pointing at a chart. The deliverable is a block of text. You copy it, you paste it, and you are done forever.

You drop the block into Claude, type “make me a skill from this,” and Claude does the Claude thing. From that moment on, every draft you feed it gets sandblasted clean of the obvious AI fingerprints before your fingers ever hover over publish.

I have been running this prompt against my own newsletter and my own audience for the past year, on articles I actually cared about. It works. The work is finished. The prompt is sitting in a code block six inches below this paragraph for less than what a Tuesday morning oat milk latte costs.

So if you are using AI to draft and you are STILL reading this without upgrading, take a quiet moment and look at the choice you are actively making with your own two hands. The cure is one click away and costs less than a coffee you will forget you drank by Wednesday. It is hiding behind a wall that exists only because you have not yet pressed a button.

You are choosing to keep publishing slop under your own name. You are choosing it on purpose. While the antidote sits one paragraph below this sentence, glowing like a quest item in a video game.

Press the button.

Read more

Useless. The kind of soft compliment a waiter gives when you order the salmon: “excellent choice”. He’s never had the salmon.

You already knew the hook was decent. You already suspected the middle was soft. The AI confirmed your priors and called it feedback.

The model has NO discipline loaded. It has no worldview to apply to your draft. So it generates the average opinion of the average reader. Polite. Vague. Wrong in a direction you can’t quite name.

This article hands you the fix.

A four-step setup. Free. Works in any AI chat. Two minutes the first time, fifteen seconds every time after that. Once it’s in place, the model stops complimenting moves you already made and starts asking the questions David Ogilvy would ask. Or Eugene Schwartz. Or Robert Cialdini.

If you don’t know those names, keep reading. That’s also fixable.

Context beats prompting

Load the model with a specific discipline before you ask it anything. That single move outperforms every prompt-engineering trick you’ve ever tried.

Once you do, the model has a worldview to apply. It asks the questions Schwartz would ask. Uses Schwartz’s vocabulary. Flags the moves Schwartz actually cared about.

The model still works like a model. The worldview is what changed. It now applies real frameworks to your draft, in the actual language of the discipline. Tested moves. Named principles. Vocabulary that maps to a body of work.

The thing you paste in is called a context profile.

The six things in every real profile

A structured set of instructions you paste into a fresh chat before you ask anything else. Or you save it as a /skill or a project. JSON usually (JSON is nothing to fear, just copy+paste). Markdown sometimes. The format matters less than what’s inside.

A real profile contains six things:

• The expert’s central thesis. What they actually believe about the work.

• The questions they ask before answering. The diagnostic moves, not the advice.

• The operating beliefs. Declarative statements they’d actually write, in their words.

• The named frameworks. Real ones, with the components broken out.

• The vocabulary. The words they use, the way they use them.

• The biases and limitations. The places where their thinking breaks down.

Miss any of those six and what you have is decoration. SKIP it.

Four steps to load a profile in two minutes

1. Find a profile. Open-source, or grab one from the Expert Profile Library.

2. Open a fresh chat. Claude or ChatGPT. Both work. If you know me, you know which one I’d pick.

3. Paste this line, then the full profile. The line is: “Use the following expert profile to evaluate, advise, and respond from the perspective of these published frameworks.” Or turn it into a skill: “Use the following information and make an expert context profile skill.”

4. Paste your draft. Ask what’s broken. Ask what the expert would change about the headline.

You’ll see the change in the first reply.

Stack profiles. Run the same draft through several of them in separate chats and watch each one surface a different failure.

• Schwartz flags whether you wrote to the right awareness stage.

• Cialdini flags which influence levers you skipped.

• Ogilvy flags whether your headline does eighty percent of the work like he said it should.

• Dunford flags whether your positioning survives five seconds in front of a stranger.

None of them tell you the writing is STRONG.

Fifteen profiles, picked and rebuilt

You can sort through open-source AI repos for real frameworks yourself. Some are gold. Most are AI-generated slop in a clean cover. Sorting one from the other takes weeks. Reading the source. Cross-checking against the original books. Verifying the frameworks behave correctly when you paste them into a chat. Restructuring the survivors into a format that loads cleanly.

I did that work over the past few weeks.

Fifteen profiles. Picked from a much larger pile. Rebuilt into a consistent schema so they all behave the same way when you paste them in.

A few of the names inside:

• Eugene Schwartz on awareness stages

• David Ogilvy on headlines

• Gary Halbert on direct response

• Robert Cialdini on the seven principles of influence

• Donald Miller on StoryBrand

• April Dunford on positioning

• Clayton Christensen on Jobs to Be Done

Plus eight more.

The free index page lists all fifteen names with a one-line use case for each. So you know which profile to grab for which problem instead of guessing.

Paid subscribers get the full JSON (simple copy+paste) for every profile, in the schema designed to load cleanly into Claude, ChatGPT, or a custom skill. You also get the operating beliefs broken out so the model knows what the expert actually believes, the vocabulary so it uses the right words, the limitations so you know when the framework will mislead you, and the source citations so you can verify any of it against the original work.

If you write for a living and you’ve ever closed an AI chat thinking “that was fifteen minutes I’ll never get back,” the library is built for you.

Browse the index here.

Ryan

The benefit is feedback that is grounded instead of generic. Most AI feedback on your writing reads like a polite editor who has never sold anything. Loaded with one of these profiles, the model asks the questions the expert would ask, applies the frameworks they tested for decades, and tells you what is broken in language that maps to a real discipline.

Writing voice and copy craft

Eugene Schwartz

The most influential copy thinker of the twentieth century. Schwartz’s central insight is that prospects exist on a spectrum of awareness about their problem, the available solutions, and your specific product. Copy that ignores where the reader actually is fails, no matter how clever the writing. The profile gives you the five stages of awareness, the five stages of market sophistication, and the headline-out method. Use it when you cannot figure out why your essay is not landing.

David Ogilvy

The man who built modern advertising on the conviction that the customer is not a moron. Specific facts beat clever adjectives. Long copy outsells short copy when you have something worth saying. The headline does eighty percent of the work. Use this profile when your writing has gotten flowery and you need a disciplined editor who punishes vague claims and rewards plain, factual specificity.

Gary Halbert

The most quoted American direct-response copywriter of the late twentieth century. Halbert’s rule of priority: the mailing list comes first, the offer second, the copy third. The voice is plain, urgent, slightly profane, and impossible to confuse with anyone else. Use this profile when you need a sales letter, a launch email, or any piece of copy that has to convert and not just charm.

Donald Miller

StoryBrand reframed marketing as a story problem. Make the customer the hero. Position yourself as the guide. Identify the problem they want solved, the plan you offer, the call to action, and the success or failure that follows. Use this profile when your writing centers itself instead of the reader, or when your offer is technically correct but emotionally flat.

Persuasion and influence

Robert Cialdini

The science of why people say yes. Cialdini’s seven principles: reciprocity, commitment and consistency, social proof, authority, liking, scarcity, and unity. He spent decades testing each one. Use this profile to audit a CTA, a sales page, or any persuasive piece for which influence levers you actually pulled and which ones you left on the table.

Chip and Dan Heath

Made to Stick. The Heath brothers studied why some ideas survive and others die, and reverse-engineered the answer into the SUCCESs framework: simple, unexpected, concrete, credible, emotional, story. Use this profile when you have an idea worth sharing but cannot get it to land. The framework tells you which of the six elements your version is missing.

Headline Craft (Ogilvy, Caples, Halbert, Wiebe)

A compilation profile pulling headline rules and tested formulas from four of the most decorated writers in direct response history. The headline carries the rest of the piece on its back. This profile gives you a battery of named formulas (the news headline, the how-to, the curiosity gap, the targeted promise) and a process for testing your own headline against them. Use it on every essay before you publish.

Positioning and audience

April Dunford

Obviously Awesome. Dunford’s argument is that most positioning is mush because it tries to be all things to all people. Strong positioning has five components: competitive alternatives, unique attributes, value, target customer, and market category. Use this profile when you cannot articulate why someone would pick your newsletter, your product, or your idea over the alternatives.

Clayton Christensen

Jobs to Be Done. Christensen reframed buying as hiring: people don’t buy products, they hire them to make progress in their lives. The four forces of progress (push, pull, anxiety, habit) explain every purchase decision. Use this profile to figure out what job your reader is actually hiring you to do, which is rarely the job you assumed.

Seth Godin (Purple Cow)

Remarkable beats average every time. Godin’s argument is that the middle is the most dangerous place to be. If your work is not worth talking about, no marketing tactic rescues it. Use this profile when you suspect your essay or your product is competent but forgettable, and you need a clear-eyed read on whether it is actually remarkable.

Seth Godin (Permission Marketing)

Anticipated, personal, relevant messages to people who chose to hear from you. Godin distinguished permission from interruption decades before email overwhelm became the default. Use this profile when you are tempted to push, pressure, or batch-blast, and need a reminder of what permission-respecting communication actually looks like.

Audience Research

A multi-method framework for actually understanding who you write for. Combines interview techniques, survey methodology, and behavioral observation from the leading researchers in the field. Use this profile before you write a launch sequence or a sales page, when you realize you are guessing about your reader instead of knowing them.

Thinking frameworks

First Principles Thinking

Break a problem down to what is actually true and reason up from there. Useful when conventional wisdom is wrong or stale, which is most of the time in fast-moving fields. Use this profile when your industry’s accepted answer feels off and you want a partner that will refuse to let you reason from analogy.

Inversion (Charlie Munger)

Munger’s mental model. Instead of asking how to succeed, ask how to fail and avoid those things. The negative path often surfaces problems the positive path missed. Use this profile to pressure-test an essay, an offer, or a decision by asking what would have to be true for it to fail catastrophically.

Cognitive Biases (Kahneman, Tversky, Thaler)

The behavioral economics audit. Forty years of research on the systematic mistakes human reasoning makes, condensed into a checkable list. Use this profile to spot the biases distorting your essay, your offer, or your CTA before they distort your reader. Anchoring, availability, confirmation, sunk cost, and the rest, applied to your specific draft.

How to use the library

Pick one profile that matches what you are working on. Open it, copy the JSON, paste into Claude or ChatGPT, paste your work. Ask for feedback. Then try a different profile on the same piece. The lenses stack. Schwartz tells you if you wrote to the right awareness stage. Cialdini tells you which influence principles you used or missed. Ogilvy tells you whether your headline carries its weight.

The library grows over time. Bookmark this page.

Who this is for: Newsletter writers and solo creators who want to understand why a reader skipped a perfectly good argument, and what to change so they read the next one.

Daniel Kahneman won a Nobel Prize for an economics paper he wrote with Amos Tversky in 1979. Neither of them was an economist. They were psychologists, and the paper described how real humans actually make decisions under risk. The answer was that we do not weigh probabilities the way the textbook said. We use shortcuts. Those shortcuts produce predictable errors. The errors are now an entire field.

Richard Thaler then turned the field into policy with Nudge. Dan Ariely turned it into a bestseller with Predictably Irrational. The whole literature has the same takeaway: readers are not rational. They are predictable. If you write as if they are rational, you will lose them. If you write as if they are predictable, you can meet them where they actually live.

This post gives you that body of work as a context profile. Drop the JSON below into Claude or ChatGPT and ask the model to audit your draft, your CTA, or your offer through Kahneman’s lens. The model will name the biases shaping the reader’s response and tell you where the writing is fighting the reader’s actual decision process.

What you get

Eight frameworks that organize the most load-bearing biases for newsletter writers: System 1 and System 2, Loss Aversion, Anchoring and Adjustment, the Decoy Effect, Social Proof, Cognitive Overload and Choice Architecture, Status Quo Bias, and Availability and Vividness. Plus nine operating beliefs, twelve vocabulary terms used the way the original researchers used them, the framework’s own biases (yes, the bias profile has a bias section), and the limitations that tell you when the lens stops being useful.

The thesis

Human judgment runs on predictable shortcuts, and those shortcuts produce predictable errors. Readers do not weigh your argument the way a rational agent would. They process it through loss aversion, anchoring, social proof, and a small set of other biases that determine whether the argument lands or slides off. Writing that ignores how the mind actually decides will be ignored, no matter how true it is.

Why this matters for newsletter writers

Most writing on persuasion treats the reader as a rational evaluator who needs better evidence. They are not. They are a person, scrolling, who decides in two seconds whether your piece earns the next two minutes. That decision is mostly System 1, mostly emotional, and mostly driven by whether the opening anchors them, whether the framing names a loss they care about, and whether the social context makes the argument feel safe to accept.

The frameworks below give you a way to audit your own work for the rhetorical traps that make writing feel hollow: the gain frame the reader cannot feel, the anchor that makes the rest of your piece sound reasonable but small, the four parallel CTAs that produce no clicks, the testimonial that feels generic. None of these are voice problems. They are decision-architecture problems. Once you can name them, you can fix them.

Preview: Loss Aversion

The single most replicated finding in behavioral economics. Losses are weighted roughly twice as heavily as equivalent gains. The same fact, framed as a loss avoided rather than a gain achieved, produces measurably stronger response.

This is not a copywriting trick. It is a structural feature of how readers process value. The reader who shrugs at “save five hours a week” pays attention to “stop losing five hours a week.” Same five hours. Different valence. Different result.

Two practical moves. First, audit every promise in your piece for whether it is framed as gain or loss, and ask whether the reader’s actual emotional state fits the frame. Anxious readers respond to loss frames; aspirational readers respond to gain frames; most newsletters mix the two and dilute both. Second, pay attention to action prompts specifically. CTAs framed as gain (”get the toolkit”) underperform CTAs framed as loss (”don’t keep losing the hours”) for tired, end-of-piece readers, who are the readers actually deciding whether to click.

A warning. Loss framing is potent and exhausting. One per piece. Stack three and the reader feels manipulated, and the inversion is permanent.

That is one of eight frameworks in the full profile. The other seven, plus the operating beliefs, the vocabulary, the limitations, and the JSON you can paste into Claude, are below for paid subscribers.

Read more

Who this is for: Newsletter writers and solo creators who want a reliable way to find the holes in their own arguments before a reader does, without becoming pessimists about their own work.

Charlie Munger built his career on a habit borrowed from a nineteenth-century mathematician. Carl Jacobi told his students that hard problems should be attacked the same way you tie a shoelace: backwards. Munger spent sixty years applying that to capital allocation, decision making, and writing, and produced one of the longest unbroken records of not being stupid in modern business history.

The Stoics got there two thousand years earlier. Marcus Aurelius rehearsed losses every morning so the actual loss, when it came, could not capsize his judgment. Same move. Different costume.

This post gives you that move as a context profile. Drop the JSON below into Claude or ChatGPT and ask the model to invert your draft, your launch plan, or your offer. The model will ask what would guarantee failure, then refuse to let you skip the answer.

What you get

Seven frameworks that organize how Munger and the Stoics thought about reversed analysis: Invert the Question, Failure Recipe, Anti-Goals, Pre-Mortem, Stoic Negative Visualization, Argument Inversion, and Decision Inversion. Plus eight operating beliefs, ten vocabulary terms used the way Munger uses them, and the limitations that tell you when inversion stops being useful.

The thesis

Most problems are easier to solve backwards. Instead of asking how to succeed, ask what would guarantee failure, then refuse to do those things. Avoiding stupidity beats chasing brilliance, because the failure paths are usually clearer, fewer, and more avoidable than the success paths are findable.

Why this matters for newsletter writers

Most writing advice is forward-only. Find your voice. Hook the reader. Build the argument. Ship it. The forward view is necessary, but it produces a predictable kind of work: confident, clean, and quietly hollow, because it never asks what would have to be true for the argument to be wrong.

Inversion forces that question. It catches the load-bearing claim that cannot survive a hostile reader, the launch plan that secretly depends on three things going right, the essay that sounds finished but collapses on the first push. Apply it once a week to your own drafts and your conversion stops looking like luck.

Preview: Argument Inversion

The version of inversion that matters most for writers. Before you publish, identify the single claim your essay is built on. The one sentence that, if a reader rejects it, makes the rest of the piece pointless.

Now write the strongest possible version of the counter-claim. Not a strawman. The smartest, best-resourced reader who disagrees with you. What would they say?

List the evidence they would marshal. Not the evidence you wish they had. The evidence they would actually have.

Now decide whether your essay addresses that evidence or evades it. Most drafts evade. The reader can feel the evasion even when they cannot name it. That is the texture of writing that sounds confident but does not land.

Revise until your argument survives its inversion. Then you have something worth publishing.

That is one of seven frameworks in the full profile. The other six, plus the operating beliefs, the vocabulary, the limitations, and the JSON you can paste into Claude, are below for paid subscribers.

Read more

Who this is for: Newsletter writers and solo creators who keep running into industry advice that does not work for them and want a structured way to figure out which rules to follow and which to ignore.

Most advice in any creative field is reasoning by analogy. People do what other successful people did. Sometimes it works. Often it produces a slightly worse copy of the original because the conditions that made the original work no longer exist. First principles is the deliberate alternative. You break the problem down to the things that are actually true, then build back up from there. Aristotle described the method in 350 BC. Descartes formalized doubt in the 1600s. Feynman used it to teach physics. Musk used it to drop rocket costs by an order of magnitude. The method is older than the internet and works just as well at a desk as it does in a factory.

This post gives you first principles thinking as a context profile drawing on its full lineage. Drop the JSON below into Claude or ChatGPT and ask the model to help you audit your assumptions, decompose a problem, or stress test a plan you are about to commit to.

What you get

Eight frameworks that organize how first principles thinkers actually work: the Three-Step Process, Cartesian Doubt, the Physics versus Convention Test, Atomic Decomposition, the Feynman Test for Understanding, the Assumption Audit Grid, the Theoretical Minimum, and the Convention Reentry Check. Plus eight operating beliefs, twelve vocabulary terms used the way the discipline uses them, and the limits where the method breaks.

The thesis

Most of what looks like a fixed truth in any field is actually a convention inherited from the way the field grew up. First principles thinking is the practice of breaking a problem down to the things that are actually true (physics, math, biology, the work that genuinely has to happen) and rebuilding from there. Reasoning by analogy copies what others have done. Reasoning from first principles produces conclusions others cannot reach.

Why this matters for newsletter writers

Newsletter advice is full of conventions disguised as laws. You have to send weekly. You have to use a hook. You have to grow on social. You have to sell a course. None of those are physics. They are conventions inherited from the writers who happened to grow first under specific conditions. First principles gives you a way to take any rule you keep tripping over and ask whether it is actually required or just borrowed. The answer often unlocks a different way of running your publication that fits you better than the templated version ever could.

Preview: The Physics versus Convention Test

Musk popularized this diagnostic but the move is ancient. When you face a constraint, classify it. Either physics, math, or biology forbids it, or only history and custom forbid it. The first kind is real. The second kind is negotiable.

Apply it to publishing schedule. Is there a law of physics that says newsletters have to ship weekly? No. The convention exists because email tools were built around weekly cadences and most successful early newsletters used that pace. Apply it to length. Is there a law that says posts have to be under twelve hundred words? No. Apply it to the paywall structure, the topic mix, the format. Almost every rule that feels like a constraint is a convention in disguise.

The discipline is to do the classification explicitly. Once you have a list of conventions you have been treating as laws, you can choose which ones to keep on purpose and which ones to drop.

That is one of eight frameworks in the full profile. The other seven, plus the operating beliefs, the vocabulary, the limits, and the JSON you can paste into Claude, are below for paid subscribers.

Read more

Who this is for: Newsletter writers and solo creators who suspect their audience is fuzzier than it should be and want a structured way to fix that.

Most newsletter writers describe their audience the same way: a job title, an age range, maybe an industry. That description does almost no work when you sit down to write. It does not tell you what the reader is actually trying to solve, what they already believe, what words they use, or where else they pay attention. The writers who close that gap convert more readers into subscribers, retain them longer, and write faster because they stop guessing.

This post gives you audience research as a context profile. It draws on the canonical thinkers of the discipline, including Indi Young, Steve Portigal, Erika Hall, Bob Moesta, Rand Fishkin, and Caroline Jarrett. Drop the JSON below into Claude or ChatGPT and ask the model to help you audit your assumptions, design a research method, or interpret the data you already have.

What you get

Eight frameworks that organize the practice of audience research: Listening Sessions, Contextual Inquiry, Just Enough Research, Jobs to Be Done Interviews, Behavioral Affinity Mapping, Survey Triangulation, Reader Reply Mining, and the Assumption Audit. Plus ten operating beliefs, twelve vocabulary terms used the way the field uses them, and the limits where each method breaks.

The thesis

You cannot write well for an audience you have not actually studied. Demographics describe what people look like. Behavior describes what they do, what they say, where they pay attention, and why they show up. Real audience research uses multiple methods to map the gap between who you think you are writing for and who is actually reading. The writer who closes that gap converts more, retains longer, and writes faster.

Why this matters for newsletter writers

Most newsletter advice assumes you already know your reader. The advice tells you how to hook them, how to pace, how to close. None of that works if the picture in your head is wrong. Audience research is the layer underneath. Get it right and the writing decisions get faster because you are no longer guessing what the reader cares about. Get it wrong and you are producing high-craft work for an imaginary person.

Preview: Listening Sessions

Indi Young’s method is the cleanest entry point into qualitative audience research. A listening session is a one-on-one conversation focused on a single recent situation the participant has lived through. The interviewer’s job is to follow the participant’s narrative, capture inner thinking and reactions, and resist every temptation to pitch, hypothesize, or steer.

The method works because people cannot reliably predict what they will do, but they can describe what they have already done. Asking “would you read a newsletter about X” gets you fiction. Asking “walk me through the last time you tried to figure out X” gets you data.

Five or six sessions with engaged readers will tell you more about your audience than any survey ever has. The output is verbatim language, real situations, and emotional patterns you can quote in your next draft.

That is one of eight frameworks in the full profile. The other seven, plus the operating beliefs, the vocabulary, the limits, and the JSON you can paste into Claude, are below for paid subscribers.

Read more

Who this is for: Newsletter writers and solo creators who want their email list to actually want to hear from them, not just tolerate them.

Seth Godin published Permission Marketing in 1999 and almost everything that has happened to email since then proves him right. The brands that earned the privilege of showing up in someone’s inbox built durable assets. The brands that bought their way in are now fighting deliverability filters and rising unsubscribe rates. If you write a newsletter, you are operating inside Godin’s framework whether you have read him or not.

This post gives you Godin as a context profile. Drop the JSON below into Claude or ChatGPT and ask the model to audit your list-building strategy, your welcome flow, or the email you are about to schedule. The model will ask the questions Godin would ask and apply the rules he laid out.

What you get

Seven frameworks that organize how Godin thinks about earned attention: the Three Requirements, the Five Levels of Permission, the Permission Ladder, the Anticipation Test, the Five Rules of Permission, the Smallest Viable Audience, and Earned Interruption. Plus ten operating beliefs, twelve vocabulary terms used the way Godin uses them, and the limits of where the framework breaks.

The thesis

Attention is the asset. Marketing that interrupts buys diminishing returns. Marketing that earns the privilege of delivering anticipated, personal, and relevant messages compounds. The job is not to broadcast louder. It is to become someone whose absence would be noticed by a defined group of people.

Why this matters for newsletter writers

Most newsletter advice is about growth tactics: hooks, lead magnets, cross-promotions, threads. Godin operates one layer underneath all of that. He is asking whether the relationship you are building is real or transactional. A list of ten thousand people who would not notice if you stopped sending is worth less than a list of eight hundred who would. The framework gives you a way to tell the difference and a way to design for the right kind.

Preview: The Three Requirements

Godin’s most cited test. Every email you send has to clear three bars at once. Miss one and the message slides back into interruption regardless of how the address was acquired.

Anticipated. The reader is actively looking forward to it. They would notice its absence. Anticipation is the load-bearing requirement. Without it, the other two do not save you.

Personal. It feels like it was written for this individual, not blasted to a generic segment. Personalization is not the merge tag. It is whether the content fits the person.

Relevant. It is about something the reader actually cares about right now. Not what you wanted to write. What they showed up to receive.

The single fastest way to lose permission is to send an email that fails any of these three tests. Most newsletters fail the anticipation test by week six. The discipline is to keep asking before you hit send.

That is one of seven frameworks in the full profile. The other six, plus the operating beliefs, the vocabulary, the limits, and the JSON you can paste into Claude, are below for paid subscribers.

Read more

Who this is for: Newsletter writers and solo creators stuck in a crowded category, watching subscribers stagnate while output stays high, and wondering why nothing is spreading.

Seth Godin published Purple Cow in 2003 and the title has been quoted by enough people who never opened the book that the original argument has gone soft in the public memory. The book is not really about cows. It is about the collapse of the old marketing model, where average products were sold to average people through paid mass attention, and what replaces it. Godin’s answer is uncomfortable for most operators. The product itself has to be the marketing. If it is not remarkable, no amount of clever copy or paid distribution will fix it.

This post gives you Godin (Purple Cow era) as a context profile. Drop the JSON below into Claude or ChatGPT and the model will look at your work the way Godin looks at a launch. It will ask whether anyone would actually miss your newsletter if it stopped tomorrow, where the safe middle is, and which edge you are unwilling to push to.

What you get

Six frameworks that organize how Godin thinks about remarkability: the Purple Cow Principle, Edges Versus the Middle, Sneezers and the Adoption Curve, Otaku, Word of Mouth Design, and the Purple Cow Lifecycle. Plus nine operating beliefs, twelve vocabulary terms used the way he uses them, and the limitations that tell you where the framework stops working.

The thesis

In a crowded market, the safe middle is invisible. Average products marketed to average people no longer travel. The only reliable engine for growth is to build a product so remarkable that people who care about it tell other people who care about it. Remarkable beats better. Edge beats middle. The product itself is the marketing.

Why this matters for newsletter writers

Most newsletter advice optimizes the wrong layer. Better hooks, tighter subject lines, more consistent posting, smarter promo. All of it useful. None of it solves the underlying problem if the work itself is not remarkable. Godin gives you a brutal but useful test. Stop a stranger in your category and ask them to describe your newsletter in one sentence. If they cannot, you are in the safe middle. If they can, but the sentence sounds like fifteen other newsletters, you are in the safe middle. The middle is where work goes to die. Godin’s framework forces the question of where you are willing to go to the edge, who you are willing to lose by going there, and what part of your work would actually move on its own if you stopped pushing it.

Preview: Edges Versus the Middle

Godin’s most operational diagnostic. Most newsletters compete in the safe middle on every dimension at once. The framework forces you to pick at least one dimension where you will go to an extreme.

Price. Free or premium beats competitive. A free newsletter has to be the most generous in the category. A paid newsletter has to be priced like it is worth the price.

Features. Radically simple or radically comprehensive beats average. One clear thing every week, or the most complete reference in the category. The middle is where most newsletters die.

Design. Distinctive beats professional-but-expected. Most newsletters look like every other newsletter. The few that look like nothing else get remembered.

Service. Extreme personal or fully self-service beats adequate. Reply to every email yourself, or build something that runs without you. The middle is the worst of both.

Speed. Instant or worth-the-wait beats same-day. Daily breaking analysis or a once-a-month essay people print out. The middle is where readers forget you exist.

Audience. A tight niche beats trying to serve everyone. Pick the smallest audience you would still be happy serving and write only for them.

Personality. Bold and human beats corporate-neutral. The newsletters that travel are written by someone, not by some.

You do not need to push every dimension. You need to push at least one until people who do not care notice and the people who care most cannot stop talking about it.

That is one of six frameworks in the full profile. The others, plus the operating beliefs, the vocabulary, the limitations, and the JSON you can paste into Claude, are below for paid subscribers.

Read more

Who this is for: Newsletter writers and solo creators who want to stop guessing why subscribers signed up, why they unsubscribed, and what would actually make them upgrade to paid.

Clayton Christensen taught at Harvard Business School for thirty years and reshaped how serious operators think about innovation. He is best known for The Innovator’s Dilemma and the disruptive innovation theory it introduced. The framework he developed late in his career, Jobs to Be Done, has quietly become the most useful customer research tool in modern product work. Most people who write a newsletter have never read him. The ones who have, run their interviews differently.

This post gives you Christensen as a context profile. Drop the JSON below into Claude or ChatGPT and ask the model to interview you about your readers, evaluate your subscriber survey, or rewrite your about page through his lens. It will ask the questions Christensen would ask and apply the framework he built.

What you get

Six frameworks that organize how Christensen thought about why people buy: the Three Dimensions of a Job, the Forces of Progress, the Universal Job Map, the Job Story Format, the Milkshake Method, and Disruptive Innovation Theory. Plus eight operating beliefs, twelve vocabulary terms used the way he used them, and the limitations that tell you where the framework fails.

The thesis

People do not buy products. They hire products to make progress in their lives. Demographic data, feature lists, and category boundaries hide more than they reveal. The unit of analysis that actually predicts buying behavior is the job the customer is trying to get done, with its functional, emotional, and social dimensions, and the forces that push them toward change or hold them back. Understand the job, and innovation becomes predictable.

Why this matters for newsletter writers

Most newsletter advice tells you to find your niche or know your audience. Christensen tells you that “audience” is the wrong unit. The same person hires different newsletters for different jobs depending on the day, the situation, and the emotional state they are in. A reader does not subscribe to your work because they are a “marketer aged 35 to 45.” They subscribe because at the moment they hit the button, they were trying to make some kind of progress. Maybe they wanted to feel less behind. Maybe they wanted to look smart in a meeting. Maybe they were avoiding a different task. The job behind the subscribe is the thing you are actually competing for, and the job behind the unsubscribe is the thing you failed to deliver. Once you can name those two jobs, every other content decision gets sharper.

Preview: The Three Dimensions of a Job

Christensen’s most underused idea. Every job has three dimensions, and most products only address one.

Functional. The practical task. For a newsletter, this might be “help me stay current on AI.” Easy to articulate. Easy to copy. Worth almost nothing on its own once the category gets crowded.

Emotional. How the reader wants to feel. Calmer. Less behind. More in control. More confident. This is the dimension that creates loyalty, and the dimension most newsletters ignore.

Social. How the reader wants to be seen. As the person who knows. As the friend others ask first. As the operator who is one step ahead. Social progress is why readers forward your work without being asked.

A newsletter that nails functional alone gets read but not loved. A newsletter that hits all three gets remembered, forwarded, paid for, and missed when it goes silent. The framework gives you a checklist for whether your work delivers all three or just one.

That is one of six frameworks in the full profile. The others, plus the operating beliefs, the vocabulary, the limitations, and the JSON you can paste into Claude, are below for paid subscribers.

Read more

Who this is for: Newsletter writers and solo creators who keep getting confused reactions to what they do, who feel stuck in a crowded category, or who suspect their problem is positioning rather than product.

April Dunford spent over twenty years running marketing at B2B startups before she wrote the book most working operators now treat as the default text on positioning. Obviously Awesome came out in 2019 and the framework inside it has become the way founders talk about positioning at this point. The reason is simple. Most other positioning advice tells you to pick a tagline. Dunford tells you to pick a category, and shows you how the category choice changes everything else.

This post gives you Dunford as a context profile. Drop the JSON below into Claude or ChatGPT and the model will run your newsletter, your landing page, or your one-line description through her framework. It will ask the questions she asks, name the trap she warns about, and push back when your positioning is generic.

What you get

Six frameworks that organize how Dunford thinks about positioning: the Five Plus One Components, the Three Positioning Styles, the Cake versus Muffin Test, the Positioning Diagnostic, the Ten Step Workshop, and the Eight Step Sales Narrative. Plus eight operating beliefs, twelve vocabulary terms used the way she uses them, and the limitations that tell you where the framework runs out of road.

The thesis

Positioning is context setting. The same product placed in a different market category competes against different alternatives, faces different price expectations, and is judged on different features. Most positioning failures are not product failures. They are context failures. Choose the category that makes your unique attributes look like must-have features and the rest of the work gets easier.

Why this matters for newsletter writers

Newsletter writers live and die by how readers describe their work to other readers. If a subscriber cannot finish the sentence “you should read this newsletter because it...” in one breath, the referral does not happen. That is a positioning problem, not a writing problem. Dunford gives you a way to choose the frame of reference your newsletter sits inside, identify what you do that the alternatives do not, and tie it to a value the right reader actually wants. Once that is locked in, the about page writes itself, the welcome email gets shorter, and the paid offer gets clearer.

Preview: The Five Plus One Components

Dunford’s core framework. Positioning is built from five linked components plus one optional element, worked in order because each depends on the one before it.

Competitive alternatives. Not who you wish you competed with. What your reader would actually do if your newsletter did not exist. Often the answer is “scroll Twitter,” “read the same five Substacks,” or “do nothing.” That is your real frame.

Unique attributes. What you have that those alternatives do not. Has to be provable, not stylistic. “Better writing” does not count. “Weekly teardown of one Substack growth experiment” counts.

Value. What those attributes let the reader do that they could not do before. Translates feature into outcome.

Target market characteristics. Who cares disproportionately about that value, and what situation they are in when they care.

Market category. The frame of reference that makes your unique attributes look obvious. The single highest-leverage decision in the whole framework.

Plus one: trends. Optional. A relevant trend that makes this value urgent right now. Skip if forced.

The framework is sequential. You cannot pick a category before you know your alternatives. You cannot identify unique attributes without comparing to those alternatives. Most positioning advice you see online skips straight to category, which is why most positioning advice you see online does not work.

That is one of six frameworks in the full profile. The others, plus the operating beliefs, the vocabulary, the limitations, and the JSON you can paste into Claude, are below for paid subscribers.

Read more

Who this is for: Newsletter writers and solo creators who keep losing the open-rate battle, the click battle, or the social-feed battle and want a structured way to write headlines that pull the right reader in.

Headline craft is the oldest, best-tested discipline in commercial writing. John Caples started running response tests for advertisements in the 1920s. David Ogilvy built an empire on the principle that the headline is the most important sentence on the page. Gary Halbert taught a generation of copywriters to obsess over the first line and treat the rest as decoration. Joanna Wiebe carried the discipline into the modern web with conversion testing on landing pages and subject lines. Together they describe the same craft from four angles, and the patterns hold across nearly a century.

This post gives you the four of them as a unified context profile. Drop the JSON below into Claude or ChatGPT and ask the model to generate, evaluate, or rewrite your headlines through the combined lens. The model will name the formula, score the specificity, check the front-loading, and warn you when you have used a category of headline the body of your piece cannot deliver on.

What you get

Eight frameworks: How-To, Number and List, Question, News and Announcement, Benefit-First, Curiosity and Intrigue, Social Proof, and the six-criterion Headline Evaluation Test. Plus nine operating beliefs, twelve vocabulary terms, the limits of formula thinking, and the JSON ready to paste.

The thesis

The headline is the highest-leverage sentence a writer ever produces. Five times as many people read the headline as read the body, which means the headline is responsible for most of the work a piece of writing does. Strong headlines are not magic. They follow a small set of patterns refined and tested across nearly a century of direct response, from Caples and Ogilvy through Halbert and into Wiebe’s modern conversion work. Writers who master the patterns stop guessing and start choosing.

Why this matters for newsletter writers

Most newsletter advice is about voice or growth. Headline craft sits underneath both. Before any of that work pays off, your subject line has to earn the open. Your subscribe page has to earn the click. Your social hook has to earn the scroll-stop. A newsletter writer who treats the headline as an afterthought is donating most of their conversion potential to the void. The four authors in this profile spent careers measuring exactly which patterns work and which patterns waste the writer’s effort.

Preview: How-To Formulas

The oldest reliable pattern in the book. How-to headlines promise practical instruction and they work for any audience that has decided they want to learn something. Caples ran them in 1932 ads. Ogilvy used them throughout the 1960s. They still work because they map cleanly to the way readers search and scroll: I want to do something, show me how.

The basic form is “How to [achieve outcome].” That is functional but plain. The strength compounds when you add structure.

A qualifier turns a generic instruction into something the reader feels was written for them. “How to Build a 6-Figure Business (Even If You Have No Audience)” outperforms “How to Build a 6-Figure Business” because the parenthetical names the obstacle the reader actually feels.

A timeframe turns a vague promise into a measurable one. “How to Learn Spanish in 90 Days” beats “How to Learn Spanish” because it tells the reader where the effort ends. Specific numbers always outperform vague ones.

A personal proof line trades the abstract instruction for a concrete success story. “How I Built a $10M Business from My Bedroom” works because the writer is staking the credibility of the headline on a real, named outcome. The reader who clicks expects to learn the actual mechanism. If the body delivers, you have built trust as well as opens.

The how-to category is also the most overused. Most writers default to it. The discipline is to ask whether your piece is genuinely procedural. If it is news, use a news headline. If it is a transformation pitch, use benefit-first. The right formula matched to the right content does most of the work.

That is one of eight frameworks in the full profile. The other seven, plus the operating beliefs, the vocabulary, the limitations, and the JSON you can paste into Claude, are below for paid subscribers.

Read more

Who this is for: Newsletter writers and solo creators who keep watching good ideas go in one reader’s ear and out the other and want a structural way to make their writing memorable enough to be repeated.

Made to Stick came out in 2007. Chip Heath teaches at Stanford, Dan Heath has spent his career at Duke, and together they spent years reverse-engineering why some ideas survive in human memory and most do not. They studied urban legends, proverbs, ad campaigns, public health messaging, and military doctrine to find the pattern. The result is the SUCCESs framework, and almost every sticky message you can think of uses some version of it.

This post gives you the Heaths as a context profile. Drop the JSON below into Claude or ChatGPT and ask the model to evaluate your draft, your tagline, or your core message through their lens. The model will tell you what is missing, where you are stuck on the abstract end of the concreteness ladder, and which of your stories are doing more work than you realize.

What you get

Eight frameworks that map the Heaths’ system: SUCCESs at the top, then Find the Core, the Curiosity Gap, the Concreteness Ladder, the Six Sources of Credibility, One vs. Many, the Three Story Plots, and the Curse of Knowledge as the diagnostic that ties them all together. Plus ten operating beliefs, twelve vocabulary terms used the way the Heaths use them, the limitations of the framework, and the JSON ready to paste.

The thesis

Most ideas die in the gap between the speaker and the listener because the speaker has the curse of knowledge and cannot remember what it was like not to know. Sticky ideas share six traits that survive that gap: they are simple, unexpected, concrete, credible, emotional, and built around a story. Ideas that get remembered, repeated, and acted on tend to use most of these traits. Ideas that fail tend to use almost none of them.

Why this matters for newsletter writers

Most newsletter advice is about voice or hook craft. The Heaths operate one layer underneath. Before you choose a hook, before you write a sentence, you have to know whether the core idea you are trying to communicate is even built to be remembered. SUCCESs is a structural test, not a stylistic one. Once you can see which traits your draft has and which it lacks, the editing decisions stop feeling subjective and start feeling diagnostic. You stop polishing prose that was never going to stick and start fixing the thing that was actually broken.

Preview: The SUCCESs Framework

Six traits, one acronym. Most sticky ideas have most of them. Most failed ideas have almost none.

Simple. Find the one essential thing and protect it from everything else. The Heaths borrow Commander’s Intent from the Army: if your reader remembers one sentence, what must it be? Simple is not dumb. Simple is prioritized.

Unexpected. Break a pattern to grab attention, then open a curiosity gap to keep it. Surprise alone is not enough. The surprise has to point at the core idea, not away from it.

Concrete. Move down the concreteness ladder until the idea has sensory hooks the reader’s brain can grab. Sarah from Portland, not one of our users. Three rings, not quickly. JFK said put a man on the moon and bring him home, not pursue space leadership.

Credible. Give the reader something they can verify themselves. Authority works. Anti-authority works. Vivid details work. Statistics work when they are translated into something a person can picture. The strongest credibility comes from a claim the reader can test.

Emotional. People feel things for one named person, not for millions. Pick one. Give them a name, a place, a problem the reader recognizes. Identity moves people more reliably than self-interest, especially over time.

Story. The Heaths describe stories as flight simulators for the brain. They let the reader rehearse the action you want them to take and supply the motivation to take it. You almost never have to invent stories. They already exist in your work. The job is to spot them and tell them well.

The whole framework is held together by one diagnostic, the curse of knowledge, which is the reason almost every important message fails on first draft.

That is one of eight frameworks in the full profile. The other seven, plus the operating beliefs, the vocabulary, the limitations, and the JSON you can paste into Claude, are below for paid subscribers.

Read more

Who this is for: Newsletter writers and solo creators who want to convert more readers into paid subscribers without resorting to pressure tactics, fake urgency, or borrowed authority that erodes trust.

Robert Cialdini spent more than three decades studying how people get talked into things. He worked undercover in car dealerships, fundraising shops, and door-to-door sales operations to find out what actually moves human behavior. Influence: The Psychology of Persuasion came out in 1984 and almost everything that followed in conversion copywriting is a footnote to it.

This post gives you Cialdini as a context profile. Drop the JSON below into Claude or ChatGPT and ask the model to review your landing page, your subscriber pitch, or your sales email through Cialdini’s lens. The model will name the principles you are using, the ones you are missing, and the places where you have stepped over the ethical line.

What you get

Eight frameworks covering the seven principles plus the combined stack: Reciprocity, Commitment and Consistency, Social Proof, Liking, Authority, Scarcity, Unity, and the Combined Persuasion Stack. Plus nine operating beliefs, twelve vocabulary terms used the way Cialdini uses them, the biases baked into the model, and the limits of where it stops working.

The thesis

Human compliance runs on a small set of universal psychological shortcuts. People rarely evaluate persuasion attempts deliberately. They react to a few near-automatic triggers built into how the mind handles uncertainty, social information, and obligation. Marketing that works tends to use these triggers honestly. Marketing that fails ignores them, fakes them, or weaponizes them and erodes trust.

Why this matters for newsletter writers

Most newsletter advice is about voice, hook craft, or growth tactics. Cialdini operates one layer underneath all of that. Before you write your subscribe page, your free-to-paid pitch, or your renewal email, you should know which psychological lever you are pulling and whether you are pulling it honestly. Cialdini gives you that map. Once you can name the principle you are using, the conversion decisions get faster and the trust your readers have in you stays intact.

Preview: Social Proof

Cialdini’s most cited and most misused principle. In situations of uncertainty, people look at what similar others are doing to decide what they should do. The more uncertain the reader, the more weight social proof carries.

The trap is using social proof that is not credible to the specific reader. A celebrity testimonial does not move a freelance designer the way a quote from another freelance designer does. Vague claims like “thousands of happy customers” are weaker than “1,847 paid subscribers as of this morning.” Generic five-star reviews matter less than reviews from named people the reader can recognize as similar to themselves.

The strongest social proof is specific, similar, and verifiable. Numbers the reader can look up. Names the reader can search. Logos the reader recognizes from their own world. Testimonials that quote the reader’s exact problem and describe the exact outcome they want.

The most common mistake is treating social proof as decoration. A wall of logos at the bottom of a landing page does almost nothing. The same logos placed beside the specific claim they validate do real work. Cialdini’s framework tells you not just to use social proof but where in the page it has to land to actually convert.

That is one of eight frameworks in the full profile. The other seven, plus the operating beliefs, the vocabulary, the limitations, and the JSON you can paste into Claude, are below for paid subscribers.

Read more