Seven days, no reason required. Email from the address you bought with and the money goes back, usually within five business days. Two things are refunded outside that window: files that never arrived, and anything on the site that was factually wrong about what you receive.

Publish threshold · 40

Your baseline · 28

Everyone sells you ideas. Nobody hands you the machine.

Three AI systems you install once and own forever. One locks every generation to your actual voice. One finds what is outperforming in your niche and tears it down automatically, every week. One tells you which of your formats to repeat and which to retire.

See the three systems Take the free one first

One-time price · Runs inside Claude, ChatGPT and n8n
No dashboard to log into, nothing to cancel

Within limits — publish
Threshold — score 40
Outside limits — repair first

Example run: twelve drafts scored by the Generic Detector. Illustrative figures, not customer data.

OnceYou buy the system, not a seat. There is no monthly anything.

~$3/moTypical API cost to run all three on your own key. That is the whole operating cost.

YoursFiles on your drive, running in your tools. If this site vanishes, your workflow does not.

The category has three failure modes. All three are pricing decisions.

Why this exists

I spent a week reading the pricing pages, feature lists and complaint threads of seven content tools charging between $19 and $149 a month. The same three problems showed up in every one of them — and none of them are technical.

The output sounds like nobody

Most tools point at the same three models with roughly the same system prompt. The result is competent, smooth and completely anonymous — the exact tone readers have now trained themselves to scroll past.

The common complaint is not that the writing is bad. It is that it reads like it came from a machine that has never met you.Recurring theme across creator forums, 2026

Your research is metered

Several tools bill by credits — and spend those credits on looking things up, not on producing anything. You end up rationing your own curiosity to protect a monthly allowance.

Paying a subscription and still counting lookups is the clearest sign that the pricing model, not the product, is in charge.Observed across four credit-metered tools

Nothing closes the loop

Every tool shows you what worked for someone else. Almost none track whether it then worked for you. That is the difference between inspiration and evidence, and it is the entire reason this exists.

Creators routinely report spending hundreds on tools that promised to save time and instead produced more work to clean up.Recurring theme across creator forums, 2026

Not apps. Not courses. Machines you install.

The three systems

Each one is a set of files — prompt chains, an importable Claude Project and Custom GPT config, automation blueprints, a scoring base. You install them into tools you already have, and they keep running whether or not I do.

Voice Lock

Your voice, extracted and enforced.

A four-stage chain that reads 15–25 things you wrote by hand and derives a voice fingerprint: sentence rhythm, vocabulary, the opinions you repeat, the things you refuse to say. Every later generation is scored against it and rejected if it drifts.

Voice-extraction prompt chain, 4 stages
Importable Claude Project and Custom GPT config
Generic Detector: 0–100 anonymity score with a hard reject threshold
Rewrite loop that repairs drift instead of just flagging it
Model-agnostic — Claude, ChatGPT, Gemini

$39one time

Get Voice Lock

The Outlier Engine

A machine that finds the proof for you.

An automation blueprint plus an agent prompt set. Every week it pulls the content outperforming its own channel baseline in your niche and scores it into your sheet. You run the teardown agent on the one it picks — hook, tension, payoff, why it travelled. Unmetered, because it runs on your key.

Importable n8n blueprint, plus a Make.com build guide
Baseline-relative outlier formula: performance against that channel's own median
Teardown agent that reports structure, not vibes
Weekly digest template you can hand straight to a ghostwriter
Setup walkthrough — about 45 minutes, no code

$59one time

Get The Outlier Engine

The Proof Loop

Retire the formats that do not work.

The part every subscription tool leaves out. You log what you publish; an AI grader compares it against your own rolling baseline and tells you, in plain language, which hook types and formats to repeat and which to drop. Your data, not the category average.

Tracking base for Notion or Google Sheets
Grader agent that returns a repeat / retire verdict
Rolling-baseline math that survives one viral outlier
Format taxonomy so you compare like with like
30-day review agent that writes the recap for you

$29one time

Get The Proof Loop

The Full Set

All three, designed to feed each other: the Engine finds the proof, Voice Lock writes it in your voice, the Proof Loop tells you whether it landed. Buying them separately works. Buying them together is the actual product.

7-day refund, no reason required. Free updates for 12 months.

$89~~$127~~one time

Get The Full Set

Measured against what you are paying for now.

How it differs

	Subscription tools	AI Resolver
What you pay	$19–$149 every month, indefinitely	$29–$89, once
Research limits	Metered by credits spent on lookups	Your API key, your call, no cap
Whose voice	A house style trained on everyone	Locked to your own top-performing posts
Where it runs	Their dashboard, their roadmap	Your Claude, ChatGPT, n8n and Notion
Measures your results	Shows you other people's numbers	Scores your posts against your own baseline
If the company disappears	The workflow disappears with it	The files are already on your drive

The honest version, so nobody asks for a refund.

Before you buy

What this is

Systems installed into tools you already pay for — nothing new to subscribe to
Built for people already publishing, who have enough history to measure something
A one-time price, with free updates as the models change underneath
Cheap to run: a few dollars a month of API usage on your own key
Files you keep, in formats you can open without me

What this is not

Not a promise that anything will go viral. Nobody can sell you that honestly
Not useful on day one of publishing — with no history, there is no baseline to score against
Not a hosted app. You install it yourself; budget 30 to 60 minutes
Not a course, not an ebook, not a video vault. There is nothing here to sit through
Not a replacement for having something to say

The Generic Detector

Free, no purchase

Score any draft 0–100 on how much it sounds like nobody.

One prompt. Paste any AI draft into it and it scores 0–100 on how much the writing sounds like nobody in particular, then names the specific sentences giving it away. It is the smallest piece of Voice Lock, and it works on its own.

81–100Indistinguishable from default model output

61–80Could be any brand account

41–60Competent and forgettable

21–40Mostly distinctive, some filler

0–20Unmistakably one specific person

Publish below · 40

100 80 60 40 20 0 Anonymity score

What you actually receive.

Questions

What exactly do I download?

Prompt files as plain text and Markdown, an importable Claude Project and Custom GPT configuration, n8n and Make.com blueprints as JSON, a Notion template plus a Google Sheets copy, and a written setup walkthrough. Everything is a file you keep. Nothing phones home.

Do I need a paid AI subscription?

Voice Lock and The Proof Loop run on the free tiers of Claude or ChatGPT, though they are noticeably better on a paid plan. The Outlier Engine needs an API key because it runs unattended; typical cost is two to four dollars a month at normal volume.

Do I need to know how to code?

No. The automation blueprints import by pasting JSON into n8n or Make — there is no scripting involved. The longest setup is The Outlier Engine at roughly 45 minutes, most of which is connecting accounts and clicking through permission screens.

Which platforms does this work for?

All of them, because the scoring is baseline-relative rather than platform-specific. It has been built and tested against LinkedIn, YouTube, TikTok, X and email newsletters. Anywhere you can export a list of what you published and how it performed, it works.

Is this just a pack of prompts?

A prompt is one instruction. These are chains with scoring rubrics, rejection thresholds, rewrite loops and a data pipeline feeding them. The value is in the loop and the measurement, not in the wording of any single instruction — which is also why they survive a model upgrade.

What happens when the models change?

Prompts get revised and re-tested when a major model ships, and you get those updates free for twelve months from purchase. After that the files still work; you just stop receiving revisions unless you renew, which is optional and cheap.

Can I get a refund?

Seven days, no reason required — email me from the address you bought with and the money goes back. Seven rather than fourteen because a week is enough to open the files and read them, not enough to run the whole system on a month of your work and then ask for it free. Two things I refund outside that window without argument: files that never arrived, and anything on this site that turns out to be factually wrong about what you receive.