chore(backlog): B66 leaves the file, its sweep having been run and recorded - #523
Merged
Conversation
…corded The entry asked for one thing: `nemotron3:33b`'s query-optimization cell swept at a raised `turnTimeoutMs`, and the result either closing the cell or written down as having failed to. That sweep ran in #500. Five runs read 3/5 again - so the lever does not close the cell - but one of them composed its report at 114 s, a run the shipped 90-second limit would have killed. The condition is met in its second direction, and what kept the entry in the file afterwards was that nobody deleted it. The result went to `docs/llms/nvidia/nemotron3.md` rather than into the entry, and that is where it belongs: a reader choosing this model looks at the model's page, and the entry would have been a third copy of numbers that already live there and in the profile document. The page says what the sweep showed and what it did not - a report turn whose length varies by a factor of twenty on identical input, which is not a limit's fault. Three things move with the entry, which is the lockstep #519 established: the M2 section's index count goes from 30 to 29, the summary bullet leaves the open-items list in `docs/AGENT.md`, and the citation on the model's page becomes the PR that closed the work rather than a pointer into a file that no longer holds it. Nothing about the cell changed. It is still locked and still marginal, and the model's page still says so.
|
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
B66asked for one thing:nemotron3:33b's query-optimization cell swept at a raisedturnTimeoutMs, and the result either closing the cell or written down as having failed to. Thatsweep ran in #500 and the entry has outlived it by three PRs.
Five runs read 3/5 again, so the lever does not close the cell — but one of them composed its
report at 114 s, a run the shipped 90-second limit would have killed. The entry's condition is
met in its second direction, and what kept it in the file afterwards was that nobody deleted it.
#518 settled forty entries and did not reach this one.
The result went to
docs/llms/nvidia/nemotron3.mdrather than into the entry, and that is where itbelongs: a reader choosing this model looks at the model's page, and the entry would have been a
third copy of numbers that already live there and in the profile document. The page says what the
sweep showed and what it did not — a report turn whose length varies by a factor of twenty on
identical input, which is not a limit's fault.
What moves with it
Three things, which is the lockstep #519 established for a closed entry:
docs/BACKLOG.md);docs/AGENT.md;(#500)— the PR that closed the work — rather than(`docs/BACKLOG.md` B66), a pointer into a file that no longer holds it.tests/unit/backlog-structure.test.tsderives the index from the entry bodies andtests/unit/agent-documentation.test.tschecks that every cited id resolves; both pass, and arepo-wide grep for
B66now returns nothing.What this does not claim
Nothing about the cell changed.
nemotron3:33b's query-optimization cell is still locked butmarginal — 5/5, 5/5, 4/5, 4/5, and 3/5 on a rested machine — and the model's page still says so.
This PR closes the bookkeeping, not the cell.