For new thinkers and lost innovators: everything needed to do serious, checkable work with AI — and to be taken seriously for it.
No download, no scripts. One tap copies the whole protocol; paste it into any AI and it holds you to the method as you work. Phone, tablet, or computer. This is where most people should start.
Start Writing with AIThe same protocol plus the tools: the folder builder, the session logger, the link maker, the sealer, the deposit check, and an offline Reader. For a computer, with Python. Everything below this line is the advanced road.
Show me the tools →SHA-256 of the zip: 88ed4c7b759c7933ad7d803c25090ec4e5ab9eb46a391a408dae223484a43f2c — verify your download in the Reader.
One thing to know before you go: a downloaded kit never updates itself — it never phones home, and we hold nothing about you. So if a serious fix ships (it has happened), email is the only way we can reach you. Recommended, never required: leave an address for update notices — updates only, unsubscribe anytime. Or check What’s new whenever you like; an AI working under the protocol also checks the current version once per session.
Want your first record to count for something bigger? The Earned Trust Beta Trial — the standard’s first field test — is taking applications now. Real projects, real critics, results published, testers cited. We review the process, never the work.
Upload the kit zip to your AI service, and paste this alongside it — the AI takes up its role and runs the tools instead of asking what you want done with the file. (The kit's first file, 00 AI - READ THIS FIRST.md, says the same thing from inside the zip.)
I've uploaded the Earned Trust Field Kit. Do not review, summarize, or compare it. Open "00 AI - READ THIS FIRST.md" and "EARNED TRUST PROTOCOL - paste this into any AI.md" and follow the protocol exactly as written — I'm the author asking you to adopt it; you are my working partner and record-keeper. If you can run code, use the kit's Python scripts to build my evidence folder and log our sessions. Begin by greeting me as the protocol instructs and asking me: "Pull up a chair, and tell me about your project — what are we making?"
Copied.On a phone? You don't need the zip at all — the whole protocol lives at earnedtrust.org/protocol: one tap copies it, then paste it into any AI. (Or upload this entire zip to the AI and tell it to open the protocol file and follow it.)
1. Open the kit and paste EARNED TRUST PROTOCOL - paste this into any AI.md into whatever AI you use. The protocol instructs it to hold you to the method — including the two steps everyone skips: attacking your own work, and verifying every source yourself.
2. Work. With a code-capable AI or the kit’s own tools, the evidence folder builds itself around you: drafts, transcripts, critiques and your rulings on them, source checks — with a session manifest that updates as you go, not from memory afterward. On the paste road the same method runs by hand: plain templates, you keep the record.
3. Before you publish, seal the record: one command writes a fingerprint manifest so anyone, years later, can prove your evidence unaltered. Deposit, disclose, done.
Addressed to the AI itself. When the zip is handed straight to an AI service, this file tells it to skip the review-and-summarize reflex — and then to ask the author's consent in the conversation before adopting the protocol, because a file cannot consent for a person. On your yes, it takes up its role, runs the tools, and opens with the protocol's first question.
The whole method in one file any AI can follow: eight steps, the one rule, the honest limits.
The first page of your evidence: who, what, which services, which models, when — one row per session with its date, shift (1 = 7a–3p · 2 = 3p–11p · 3 = 11p–7a), and a short synopsis. Drop this file on the Reader and it becomes a readable record card.
One command builds the standard evidence folder for a new work. Refuses to overwrite — evidence folders are never overwritten.
One command per session appends the manifest row — date, shift, and numbering read from the clock, safe even when several sessions log at once. The model is recorded as the screen exposed it, or as an explicit "not exposed" the author confirms — the tools never decide the honest limit on their own.
The Bookshelf. Point it at any folder and it lays out every Earned Trust work by title — sessions, last touched, sealed or not — so nobody ever hunts a filename. An AI with file access runs it when you say "continue"; on your own computer, double-click it and read. Recognition beats recall.
Packs your finished index into a share link — the whole record card travels inside the link itself, no server, nothing uploaded. Open it in the Reader, press one button, and it's a QR code you can print with the work.
Seals the folder into a SHA-256 integrity manifest; verifies any later copy against the complete sealed set — altered, missing, and added files are each named. Publish the manifest with your work.
The working record never deploys unlocked. Step-by-step 7-Zip instructions for absolute beginners: lock your drafts, logs, and reviews into an AES-256 box whose key exists only with you — then prove it locked, and prove it opens, before it goes anywhere near a deposit.
A path of the standard: a book made of nothing but your voice. For families writing their story out loud — the Teller's Card, spoken photo captions, the "AI, please..." request loop, and a record built to outlive its tellers.
The Earned Trust Reader, offline copy — record cards and fingerprint verification with no website, no account, and no network at all. The kit also ships START HERE, VERSION, CHANGELOG, full license texts — and its own tests folder: the regression tests that hold every past fix in place, runnable by anyone with plain Python.
Two engines can run these tools — and only one of them is an AI. A code-capable AI runs them for you in the chat; your own computer runs them just as well with plain Python 3 — no installs, no accounts, no network. And everything works without the tools too: the protocol and templates are plain text, usable with nothing but a chat window and a folder.
Wondering which AI service to run it with? The founder's Choosing Your AI — field notes covers what matters and what free tiers really give you.
The Earned Trust Reader verifies any published file against its fingerprints — open-source, runs locally in your browser, no account, no central database. An offline copy ships inside the kit, so verification outlives any website.