feat(research): a gate for the one question the other paper gates cannot ask - #732
Open
gHashTag wants to merge 1 commit into
Open
feat(research): a gate for the one question the other paper gates cannot ask#732gHashTag wants to merge 1 commit into
gHashTag wants to merge 1 commit into
Conversation
…not ask Every existing paper gate asks whether the manuscript is internally consistent: that its figures resolve, its references land, its numbers agree with each other. A manuscript can satisfy all of that and describe a state of knowledge that has been superseded, and on 2026-08-24 this one did. `tnf_paper.tex` was last written 2026-08-20. `measurements/` had been written through 08-23. In between: the first silicon numbers (W973, Fmax 80.35 MHz, timing PASS), a defect the die found that simulation had passed (W975-W978), the operator table becoming derivable (W989), both spec fixes confirmed across four placements (W990), and W991 -- which shows TNF holding fewer values and a step 12x coarser than posit at matched physical width, because the ternary exponent uses 81 of 128 codes and leaves 8190 of 2^19 unreachable. None of it is in the paper. Not one occurrence of `80.35`, `BSCAN`, `on-die`, `516096` or `524286`. SUBMISSION.md's pre-flight passes every check it makes; none of them is this one. The gate reports 21 wave-labelled records newer than the manuscript and named nowhere in it -- W973 through W992, an unbroken run. Records with no wave label are reported separately and explicitly as suspicion rather than finding: the fallback there is five-digit figures, which cannot distinguish an omission from a coincidence. It judges nothing. A record deliberately left out of the paper is a decision, and the gate's own advice is to make that decision visible by naming the wave in the text even to exclude it. Dates come from `git log`, not mtime, because a checkout rewrites every mtime to now -- which would make this gate pass on any fresh clone and fail nowhere, quietly, forever.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Every paper gate here asks whether the manuscript is internally consistent — figures resolve, references land, numbers agree with each other. A manuscript can pass all of that while describing knowledge that has been superseded. On 2026-08-24 this one did.
An unbroken run W973–W992. The paper contains no occurrence of
80.35,BSCAN,on-die,516096or524286— so it carries neither the strongest new evidence (the format's operators reached silicon) nor the most damaging new refutation (at matched physical width TNF holds fewer values and steps 12× coarser than posit, because the ternary exponent uses 81 of 128 codes and leaves 8190 of 2^19 unreachable).Every pre-flight check in
SUBMISSION.mdpasses. None of them is this one.What it does not do
Judge. A record deliberately left out is a decision; the gate's advice is to make that decision visible by naming the wave in the text, even to say it is excluded.
Records with no
wavefield are reported separately and as suspicion, not finding: the fallback there is five-digit figures, which cannot tell an omission from a coincidence. 21 provable, 42 suspected — and the gate says which is which rather than reporting 63.One implementation note
Dates come from
git log, not mtime. A checkout rewrites every mtime to now, which would make this gate pass on any fresh clone and fail nowhere, quietly, forever.