WAKES WHEN a chat is long or a big file is about to be read whole, or someone asks will it fit
THE LOOP
weigh · tokens, heaviest first
AGROUND? write the ledger line first
cut · pieces at paragraph edges
read one piece at a time, in order
join · MATCH before the original goes
NEVER
Print an estimate as the AI's own count
Keep a cut that does not join back
Weigh pictures; THE HOLD CARD keeps them
IT MAKES weigh — estimated tokens, heaviest first, CLEAR, SHOALING or AGROUND; cut — numbered pieces at paragraph edges, and the hash of the whole; join — the pieces put back: MATCH or DIFFERENT.
THE STANDARD weigh before reading; edges, not middles; byte for byte; the ledger first; an estimate, said so; the other budget.
PROOF SELFTEST 8/8 · witnessed in Claude 2026-10-11 · 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 · 73A8883EF5F9B6A229FE8EB0
CITE THIS CARDTHE SOUNDING CARD, Edition 1. Lund Studio LLC, Westminster, Colorado. https://lundr.tech/?view=sounding-card. Punch 73A8883EF5F9B6A229FE8EB0. Retrieved 2026-10-11.
What it is.
Weighs what an AI is about to read and cuts what will not fit. weigh reads files and folders, estimates tokens (four characters a token for prose, three for code), names the heaviest first and gives CLEAR, SHOALING or AGROUND against a budget the person sets; cut splits a long text file at paragraph edges into numbered pieces that each fit a chunk, with the SHA-256 of the whole beside them; join puts the pieces back and proves the result is the original byte for byte, or says DIFFERENT. Pictures are set aside: they spend a different budget.
It carries a machine that shows its working: weigh — estimated tokens, heaviest first, CLEAR, SHOALING or AGROUND; cut — numbered pieces at paragraph edges, and the hash of the whole; join — the pieces put back: MATCH or DIFFERENT — 8 checks in its selftest, every command on the record. It will never print an estimate as the AI's own count; never keep a cut that does not join back; never weigh pictures; THE HOLD CARD keeps them.
Six laws, written into the card.
They are in the card itself, so the AI follows them.
WEIGH BEFORE READINGThe heaviest file is named before any file is opened.
EDGES, NOT MIDDLESPieces end at paragraph edges, then line edges; a line is cut mid-way only when it is longer than a piece.
BYTE FOR BYTEA cut is proved by join against the hash of the whole; DIFFERENT means the pieces are not trusted.
THE LEDGER FIRSTWhen the depth reads AGROUND, the ledger line is written while there is still room to write it.
AN ESTIMATE, SAID SOTokens are estimated from characters; the AI's own counter is the count.
THE OTHER BUDGETPictures are not weighed here; THE HOLD CARD keeps the image budget.
What rides in the card.
In the card’s own voice, with a machine that shows its working.
THE GUIDEreferences/THE-SOUNDING.md
THE MACHINEsounding.py · three commandsweigh — estimated tokens, heaviest first, CLEAR, SHOALING or AGROUND; cut — numbered pieces at paragraph edges, and the hash of the whole; join — the pieces put back: MATCH or DIFFERENT. 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.
weigh Estimated tokens, heaviest first, CLEAR, SHOALING or AGROUND
cut Numbered pieces at paragraph edges, and the hash of the whole
join The pieces put back: MATCH or DIFFERENT
On the record Three text files weigh about 32,900 tokens: AGROUND against a budget of 20,000 and CLEAR against 200,000, with the guide named first and the picture set aside. The guide is cut into 14 pieces, the largest about 1,833 tokens, and joins back as a MATCH; after one byte is added to a piece the join reads DIFFERENT, and a missing file is refused.
What it will never do.
It will never print an estimate as the AI's own count; never keep a cut that does not join back; never weigh pictures; THE HOLD CARD keeps them.
Tokens are estimated from characters, not counted with any AI's tokenizer, and the estimate drifts for languages other than English. It reads text files as UTF-8; pictures, PDFs and other binaries are set aside, not weighed. It cannot see how full a conversation already is: it weighs what it is handed. A rebuild of the September 2026 machine from its recorded source; the MATCH check against the hash of the whole is new in this edition. 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-SOUNDING.md — the guide in references/, in the card's own voice
scripts/sounding.py — the machine: weigh · cut · join, 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 2026-10-11 · EDITION 1 · PUNCH 73A8883EF5F9B6A229FE8EB0
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 56 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 2026-10-11
sounding.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.
weigh
estimated tokens, heaviest first, CLEAR, SHOALING or AGROUND
cut
numbered pieces at paragraph edges, and the hash of the whole
join
the pieces put back: MATCH or DIFFERENT
selftest
8 checks that prove the machine; it must print 8/8
weigh · ESTIMATED TOKENS, HEAVIEST FIRST, CLEAR, SHOALING OR AGROUND
$ python3 scripts/sounding.py weigh notes --budget 20000
THE SOUNDING · 3 file(s) · about 32,902 tokens against a budget of 20,000 tokens (165%)
~24,438 tokens · 97,749 characters · guide.md <- over the budget by itself
~8,334 tokens · 25,000 characters · tool.py
~130 tokens · 520 characters · readme.txt
AGROUND · this does not fit: write the ledger line now, then cut and read one piece at a time
not weighed: 1 file(s) that are not text (pictures spend a different budget)
weigh · ESTIMATED TOKENS, HEAVIEST FIRST, CLEAR, SHOALING OR AGROUND
$ python3 scripts/sounding.py weigh notes --budget 200000
THE SOUNDING · 3 file(s) · about 32,902 tokens against a budget of 200,000 tokens (16%)
~24,438 tokens · 97,749 characters · guide.md
~8,334 tokens · 25,000 characters · tool.py
~130 tokens · 520 characters · readme.txt
CLEAR · this fits with room left to work
not weighed: 1 file(s) that are not text (pictures spend a different budget)
cut · NUMBERED PIECES AT PARAGRAPH EDGES, AND THE HASH OF THE WHOLE
$ python3 scripts/sounding.py cut notes/guide.md --chunk 2000 --out pieces
cut guide.md into 14 piece(s) in pieces · the largest about 1,833 tokens · chunk 2,000
read them one at a time, in order; prove the cut with join before the original is set aside
join · THE PIECES PUT BACK: MATCH OR DIFFERENT
$ python3 scripts/sounding.py join pieces --out guide-back.md
joined 14 piece(s) · 97,749 bytes · sha256 781c15ea8d42374d4fb68f13
MATCH · byte for byte the file that was cut
join · THE PIECES PUT BACK: MATCH OR DIFFERENT
$ python3 scripts/sounding.py join pieces
joined 14 piece(s) · 97,750 bytes · sha256 96d212651f90e6b570bda5df
DIFFERENT · the pieces do not join back to the file that was cut (sha256 781c15ea8d42374d4fb68f13 expected)
cut · NUMBERED PIECES AT PARAGRAPH EDGES, AND THE HASH OF THE WHOLE
$ python3 scripts/sounding.py cut notes/missing.md
no such file: missing.md
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?
Weigh · tokens, heaviest first. It asks first for the files or folders about to be read, the budget in tokens, and for a cut the size of each piece.
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. Tokens are estimated from characters, not counted with any AI's tokenizer, and the estimate drifts for languages other than English. It reads text files as UTF-8; pictures, PDFs and other binaries are set aside, not weighed. It cannot see how full a conversation already is: it weighs what it is handed. A rebuild of the September 2026 machine from its recorded source; the MATCH check against the hash of the whole is new in this edition.
What will it never do?
It will never print an estimate as the AI's own count; never keep a cut that does not join back; never weigh pictures; THE HOLD CARD keeps them.
