A hygiene scan of the lab's Lean proofs
Counted in the lab’s Lean source, with comments removed first
- Lean files scanned
- 339
- Theorem and lemma declarations
- 1,861
- Toolchain files, all set to one version
- 29
A text scan of source files, not a build of every corpus.
A text scan of the lab's Lean files found no unfinished proof placeholders and no unchecked decision shortcuts in code; it counts declarations in source and does not compile every corpus
What it does not claim
It is a text scan of source files, not a clean build of every corpus; one corpus is counted but not compiled. The declaration count is one of several the lab keeps, which count different things and must not be added. The lab's code is not public yet.
The limits, in the lab’s words:
AST-F5-02's own registry status reads 'not built in Pass A or Pass B' — the transport corpus is COUNTED, NOT COMPILED.
STILL OPEN, per the grader: this is a textual scan of tracked sources, not a clean elaboration of every corpus; the stronger corpus-wide compilation claim needs the builds.
Sentences of the lab’s limits, word for word from the published file; the rest is published by fingerprint only.
A proof checker can be asked to accept a step without proof, through an unfinished placeholder, an extra axiom or an unchecked decision shortcut. A proof that leans on one of these is not fully checked. The lab scans its Lean files for all three and counts the theorem declarations they carry.
What it shows
In this measurement: a scan of 339 tracked Lean files, with comments stripped first, across 29 toolchain files, all set to the same version.
The files carry 1,861 theorem and lemma declarations, and the scan finds no unfinished proof placeholders and no unchecked decision shortcuts in code, and no extra axiom of the lab's own. The lab's exact words, which name the proof checker's keywords for these, are in the file linked below.
Corrections the lab made
An earlier version of the scan counted its own documentation and reported hits that were sentences saying there were none; it now strips comments first and plants each kind of hit to prove it can fire. An earlier declaration count included comment lines and was too high; the count above is the corrected one.
Why it matters, and to whom
For a buyer relying on the Lean-checked results on this site, the scan is a first check that the proofs do not lean on unchecked steps. It is not a substitute for compiling them.
Why now
Machine-checked proofs are increasingly offered as evidence for engineering claims, and their value depends on there being no unchecked step in them.
How it was checked
The scan strips comments, searches the code, exits with an error on any hit, and carries a self-test that plants each kind of hit and requires the scan to find it.
How to reproduce
The lab's claim, word for word, with the figures found in it and the sentences of its limits are published here; the lab's command and commit are named in the measured results. Each file is listed with its hash in the site's receipt manifest.
The result, in the lab’s exact words
Read it with this condition: the claim is a text scan of tracked source files with comments removed; it does not compile every corpus, and one corpus it counts is counted, not built.
The lab’s claim is published word for word in its own file. It names the proof checker’s own keywords, which this site’s pages do not print.
Prior art
No outside prior-art search is on record for this result yet.
Evidence
- The measured results behind this page, read from the lab’s own files at a named version.
- The lab’s current claim and limits, word for word.
- Every file this site publishes.