803 recorded inventions. Every fingerprint is public. No content is. You can check the count, the dating and the integrity in about fifteen minutes — without a key, without permission, and without any cooperation from me.
Anyone asked to assess a documented body of knowledge normally faces an unpleasant choice. Either they are shown the content — and are no longer an independent assessor but a party to it. Or they are not shown it — and must take everything on trust.
This page removes that choice. Nothing below asks you to believe me. Every step is a computation you can repeat.
doi.org/10.5281/zenodo.22909342 — a concept identifier, so it always resolves to the newest version. Open access, no account.
English and German, each as plain text and as a typeset document, together with the public register.
sealed-register.csv — also served here directly. It has 803 rows and exactly four fields:
| Field | Content |
|---|---|
entry_no | sequential number, 1 to 803 |
function | one of ten function groups, in plain words |
sealed_on | date of sealing |
sha256 | fingerprint, 64 characters |
It contains no content. No title, no description, no assessment. A SHA-256 fingerprint is one-way: the content cannot be recovered from it. That is the defining property of the function, not a claim of mine.
All 803 fingerprints, in ascending entry order, joined with newlines and hashed once more, give exactly this:
Check it. One line is enough:
python -c "import csv,hashlib;r=sorted(csv.DictReader(open('sealed-register.csv',encoding='utf-8')),key=lambda x:int(x['entry_no']));print(hashlib.sha256(chr(10).join(q['sha256'] for q in r).encode()).hexdigest())"Change one character, remove one row or add one, and the root no longer matches. The body is therefore protected against silent insertion and silent deletion. Integrity alone would not give you the second one.
verify_seal.py — about twenty lines of logic, readable in one sitting. It recomputes the fingerprint of an opened entry from its eight fields and compares it against the published one. No network, no service, no key, no permission.
python verify_seal.py --selftestThe self-test demonstrates the computation on an invented example entry. It discloses nothing real.
Why the method is published: a secret verification method proves nothing. The openness of the method is precisely what makes the sealing checkable. The protection lies not in the method but in the inputs.
doi.org/10.5281/zenodo.22909339 — access restricted. You see title, author, date and version history; not the content. That is deliberate, and it is the point. The access decision is enforced by the repository, not merely asserted by me.
The baseline of the whole body sits in Bitcoin block 945970:
Block height: 945970
Block hash: 000000000000000000002fc55b1780b87c23bcbdb1af3b7cd4329f23f71d27d5
Unix time: 1776722274 = 2026-04-20 21:57:54 UTC
Anchored digest:
299e4b3d0ac9e50740268896b5eeb541f6462332bb67b7efdf4cd309704770d0Check the block with any provider you like, and convert the timestamp yourself:
python -c "import datetime;print(datetime.datetime.fromtimestamp(1776722274,datetime.UTC))"No one can alter a Bitcoin block after the fact. What was anchored at that moment existed at that moment.
Stating this plainly is part of the point, so it is stated here rather than left to be discovered.
| Established | Not established |
|---|---|
| That 803 delimited entries exist | That every one of them is valuable |
| That they existed in exactly that form on a given date | That they are novel in the sense of a registered right |
| That nothing has since been altered, removed or inserted | How good the content is |
| That the content was not disclosed | That no one elsewhere held something similar earlier |
These are statements about form, count, time and integrity. They are not statements about quality or worth. Conflating the two would be dishonest, and it is exactly the failure mode that sinks comparable projects.
If anyone holds earlier work of this kind, the appropriate response is to date it. That is why the foundations are public rather than in a drawer.
If your assessment requires the content of a particular entry, the path is short:
verify_seal.py to confirm for yourself that they are precisely the sealed ones.You remain free to use and publish your conclusion. Statements about count, structure, dating, verifiability, methodological classification and assessability are expressly not restricted. What is protected is the content, not your judgement.
The usual choice between show it and lose it and stay silent and prove nothing does not arise.
Proof is decoupled from negotiation. You need not believe me, and I need not disclose anything for you to verify. Only when you want to judge the substance do we open exactly as much as that requires — and not one entry more.