THE VOICE CARD · your voice, from your own writing
WAKES WHEN an AI drafts in someone's name, or someone asks whether a draft sounds like them
THE LOOP
learn <samples> · the voice, measured
draft as yourself, or let the AI draft
score <draft> --voice · the match
mend the measures named furthest off
compare voices when two people write
NEVER
Learn a voice from writing not its owner's
Pass off a draft as someone who never saw it
Call a measure a judgement of quality
IT MAKES learn — the voice: its measures, the words it reaches for, how it opens; score — a match out of 100 and the measures furthest off, with their direction; compare — two voices, measure by measure.
THE STANDARD the owner's own writing only; measured, not guessed; the spread is part of the voice; off is named with its direction; a draft in someone's name is theirs to approve; style, not truth.
PROOF SELFTEST 8/8 · witnessed in Claude 2026-10-10 · Edition 1.
THE SEAL 98 of 100 · LEGENDARY · WITNESSED · Trigger 5 · Machinery 5 · Law 5 · Portability 5 · Proof 5
IT CAN BE WRONG an AI reading a text file: it can be wrong and can invent a number, a rule, a name or a date; check what matters against your own records and the source; the risk is yours. THE-CLAUSES.md rides in the packet and is part of the licence.
THE PUNCH · SHA-256 OF SKILL.MD · E0E31A619E281DF660830CA4
CITE THIS CARDTHE VOICE CARD, Edition 1. Lund Studio LLC, Westminster, Colorado. https://lundr.tech/?view=voice-card. Punch E0E31A619E281DF660830CA4. Retrieved 2026-10-10.
What it is.
Learns a writer's voice from their own writing and holds drafts to it: learn reads a folder of the owner's samples and keeps how they write — words a sentence and the spread of it, letters a word, contractions, questions, exclamations, I and you, commas, the words they reach for, how they open a sentence; score reads a draft against that voice and gives a match out of 100, naming the measures furthest off and which way (sentences run long, no contractions, no first person); compare sets two voices side by side.
It carries a machine that shows its working: learn — the voice: its measures, the words it reaches for, how it opens; score — a match out of 100 and the measures furthest off, with their direction; compare — two voices, measure by measure — 8 checks in its selftest, every command on the record. It will never learn a voice from writing not its owner's; never pass off a draft as someone who never saw it; never call a measure a judgement of quality.
Six laws, written into the card.
They are in the card itself, so the AI follows them.
THE OWNER'S OWN WRITING ONLYA voice is learned from samples its owner wrote or has the right to use; the card does not learn a living writer's voice to imitate them.
MEASURED, NOT GUESSEDSentence length and spread, word length, contractions, questions, exclamations, first and second person, commas — each counted from the samples.
THE SPREAD IS PART OF THE VOICEEach measure may drift by the samples' own spread before it counts as off.
OFF IS NAMED WITH ITS DIRECTIONThe draft is told which measures are furthest from the voice and which way: longer, shorter, more, fewer.
A DRAFT IN SOMEONE'S NAME IS THEIRS TO APPROVEThe card scores; the person whose voice it is reads and signs off.
STYLE, NOT TRUTHA draft can match the voice and still be wrong; the card reads how it is written, not what it says.
What rides in the card.
In the card’s own voice, with a machine that shows its working.
THE GUIDEreferences/THE-HORIZON.md
THE MACHINEvoice.py · three commandslearn — the voice: its measures, the words it reaches for, how it opens; score — a match out of 100 and the measures furthest off, with their direction; compare — two voices, measure by measure. Every command prints its working and refuses a bad input; 8 checks in its selftest, every command on the record.
What it does once your AI has it.
learn The voice: its measures, the words it reaches for, how it opens
score A match out of 100 and the measures furthest off, with their direction
compare Two voices, measure by measure
On the record A voice learned from three samples — short sentences, contractions, first and second person, the words the writer reaches for; a draft in that way scored IN THE VOICE; a committee paragraph scored far, with its long sentences, long words and no contractions named.
What it will never do.
It will never learn a voice from writing not its owner's; never pass off a draft as someone who never saw it; never call a measure a judgement of quality.
The measures are counts of form for English prose; two writers can share them and one writer can vary by mood and audience, so a high match is a likeness of form, not proof of authorship. Too little writing is refused rather than read: three sentences is the floor, five pieces a fair start. It is an AI reading a text file: it can be wrong, so check what matters against your own records and the source.
What’s in the buy.
CARD.md — the whole card: who it is, its laws, when it says no, what it carries · SKILL.md — the door your AI reads first · START-HERE.md — the first five minutes, for you
THE-HORIZON.md — the guide in references/, in the card's own voice
scripts/voice.py — the machine: learn · score · compare, standard library only, selftest 8/8
WITNESS.md + SELFTEST.md — every command on the record, as it printed · MANIFEST.md, every file hashed · AUDIT.md, the twelve checks
MY-LEDGER.md — the card's memory, on your own machine · TIN.md · LICENSE-NOTE.md · THE-CLAUSES.md · HOSTS.md
card.svg + card-machine.svg + card.json · card.html, the card as one offline file · the doors for ChatGPT, Gemini, Cursor, Claude Code and MCP hosts (rules/, system/, mcp/)
PROOF · SELFTEST 8/8 · 3 COMMANDS ON THE RECORD · WITNESSED IN CLAUDE 2026-10-10 · EDITION 1 · PUNCH E0E31A619E281DF660830CA4
ON EVERY DOOR · A CARD FOR YOUR AI, A FILE FOR YOUR COMPUTER
In Claude: turn on code execution under Settings › Capabilities, then Customize › Skills › + › Create skill, and upload the zip as it is — START-HERE.md in the box walks you through it, and tells you how to keep the ledger. In ChatGPT, Gemini and the rest: a Project, a GPT or a Gem with the card’s files attached and the short door in the box pasted as its instructions. The words run anywhere an AI reads Markdown; the machines run wherever the host runs Python. Witnessed in Claude only. The host marks →
Take it.
Nº 1107 PROPOSED — LAUNCHING SOON
Sold as is under the term printed. A card instructs an AI you operate; results depend on your model, your files and your judgement. Not medical, legal, tax or financial advice, never the person’s boss, and not a licence to practise. The terms →
One payment. The file is yours to keep and keeps working at its edition; new editions issued in the year after you buy reach you by email as replacement files. Tax is added at checkout where it applies. Nothing auto-renews.
A TERM, NOT A SUBSCRIPTION: NOTHING RENEWS ITSELF AND NOBODY IS CHARGED AGAIN. WHEN A TERM ENDS THE FILES STAY ON YOUR COMPUTER AND STAY READABLE — FREEZE, NOT LOSS.
THE MACHINE · IN THE OPEN · FROM THE WITNESSED RUN OF 2026-10-10
voice.py, every command showing its working.
A small Python program, standard library only. Your AI runs it; you can too. Every block below is pasted from the card’s own WITNESS.md, as printed.
learn
the voice: its measures, the words it reaches for, how it opens
score
a match out of 100 and the measures furthest off, with their direction
compare
two voices, measure by measure
selftest
8 checks that prove the machine; it must print 8/8
(three samples of the owner's own writing)
$ (three samples of the owner's own writing)
learn · THE VOICE: ITS MEASURES, THE WORDS IT REACHES FOR, HOW IT OPENS
$ python3 scripts/voice.py learn samples --out voice.json
voice learned from 3 samples · 46 sentences · 176 words
words a sentence 3.8
spread of sentence length 0.8
letters a word 3.6
contractions per 100 words 10.8
questions per 10 sentences 0.9
exclamations per 10 sentences 0.0
I and me per 100 words 8.0
you per 100 words 0.6
commas a sentence 0.2
reaches for: bake, ovens, wake, quiet, you'd, sell, noon, isn't, point, dough
opens with: i, the, we, you'd, isn't, flour
written: voice.json
(two drafts: one in the owner's way, one by a committee)
$ (two drafts: one in the owner's way, one by a committee)
score · A MATCH OUT OF 100 AND THE MEASURES FURTHEST OFF, WITH THEIR DIRECTION
$ python3 scripts/voice.py score draft-mine.md --voice voice.json
match 83 of 100 · 5 sentences read · 3 of the writer's words used
OFF · you per 100 words: 3.8 against the voice's 0.6 — higher
OFF · contractions per 100 words: 7.7 against the voice's 10.8 — lower
IN THE VOICE
score · A MATCH OUT OF 100 AND THE MEASURES FURTHEST OFF, WITH THEIR DIRECTION
$ python3 scripts/voice.py score draft-committee.md --voice voice.json
match 46 of 100 · 3 sentences read · 0 of the writer's words used
OFF · letters a word: 6.9 against the voice's 3.6 — higher
OFF · I and me per 100 words: 0.0 against the voice's 8.0 — lower
OFF · contractions per 100 words: 0.0 against the voice's 10.8 — lower
OFF · words a sentence: 17.7 against the voice's 3.8 — higher
NOT YET THE VOICE — mend the measures named above
compare · TWO VOICES, MEASURE BY MEASURE
$ python3 scripts/voice.py compare voice.json voice.json
words a sentence 3.8 3.8
spread of sentence length 0.8 0.8
letters a word 3.6 3.6
contractions per 100 words 10.8 10.8
questions per 10 sentences 0.9 0.9
exclamations per 10 sentences 0.0 0.0
I and me per 100 words 8.0 8.0
you per 100 words 0.6 0.6
commas a sentence 0.2 0.2
words both reach for: bake, don't, dough, flour, isn't, knows, noon, ovens, point, quiet, rush, sell, wake, water, you'd
THE NUMBERS, NAMES AND DATES IN THESE RUNS ARE THE STUDIO’S OWN EXAMPLES, INVENTED FOR THE RECORD; NO REAL COMPANY, PERSON OR PROPERTY IS NAMED.
QUESTIONS AT THE TABLE
Asked before the wire.
What does it need from me?
Learn <samples> · the voice, measured. It asks first for a folder of the owner's own writing — five pieces or more — and the draft to read.
Does it send anything anywhere?
No. The machine is a standard-library Python program that runs on your own computer and sends nothing anywhere; the words run wherever an AI reads Markdown. No server, no account, no key.
What is on the record?
The self-test (8/8) and every command run on real example inputs, pasted into WITNESS.md as they printed; the punch on the face is the SHA-256 of SKILL.md, so a downloaded copy can be checked against this room.
Can it be wrong?
Yes — it is an AI reading a text file. The measures are counts of form for English prose; two writers can share them and one writer can vary by mood and audience, so a high match is a likeness of form, not proof of authorship. Too little writing is refused rather than read: three sentences is the floor, five pieces a fair start.
What will it never do?
It will never learn a voice from writing not its owner's; never pass off a draft as someone who never saw it; never call a measure a judgement of quality.
