Skip to content
This repository was archived by the owner on Nov 21, 2026. It is now read-only.

docs: method text for the 2026-11-03 amendments - #561

Merged
tae898 merged 1 commit into
mainfrom
docs/method-2611-amendments
Nov 3, 2026
Merged

tae898 merged 1 commit into
mainfrom
docs/method-2611-amendments

Conversation

@tae898

@tae898 tae898 commented Nov 3, 2026

Copy link
Copy Markdown

Closes #560

Docs only, three files under benchmarks/experiments/. No code, rows or stage scripts change, so the mini (which pulls main at each stage start) executes the same thing before and after. Not merged on purpose: for the owner to merge.

file was stale now says
CAMPAIGN.md the paper profile: experiments must finish by 2026-11-10 numbers land about 2026-11-15; submission 2026-11-24 (SIGMOD 2027 industrial track) or 2026-12-01 (ICDE 2027 industry track, fallback); bench host order after qC48 is the re-runs (qC14 at turbo off, qC15..qC23 at the 10 us hold), the four-arm ArcadeDB pass on the released 26.11.1 wheels, then passes 4 and 5 (qC49..qC80, PASS45.go); comparators are N=3 for sure, N=5 where passes 4 and 5 finish before the tables freeze
PROTOCOL.md section 2 nothing about excluded windows or the hold-0 relaunch a bullet listing the windows (qC14 turbo, 2026-11-01T05:26:21Z to the start of qC24 at hold 0 us, the 65 s turbo flip in qC22 on 2026-11-03) and the dense large-tier rep 1 (qC06) that is not a speed reference
PROTOCOL.md section 7 the idle-hold override row said an unset launcher measures at hold 0 adds that it happened once (the 2026-11-01 relaunch bypassed launch.sh), how it is handled, and that an unset hold is to become a hard error after the pass
PAGE-SPEC.md the turbo paragraph said a stage refuses to start with turbo on adds that the check runs only at a stage's start and turbo came back on twice during the pass

Sources: DECISIONS #228, #232 and its amendments, #234, #235; repros/turbo-audit-20261103; repros/dense-rep1-20261103. The window for 2026-11-01T05:26:21Z to the start of qC24 is still to be entered in excluded_windows.json when qC24 starts (OPEN-ITEMS 102), which is why the text says so instead of naming an end time. structure_check.py prints the same output before and after the edit.

🤖 Generated with Claude Code

…cluded windows, run order, paper dates)

CAMPAIGN.md: the paper profile now says numbers about 2026-11-15, submission 2026-11-24 (SIGMOD industrial) or 2026-12-01 (ICDE industry, fallback), and the bench host order (re-runs, four-arm pass, then passes 4 and 5). PROTOCOL.md: a bullet for the excluded windows and the dense qC06 rep 1, and the hold-0 relaunch in the idle-hold override row. PAGE-SPEC.md: the turbo gate checks at a stage start only. Source: DECISIONS #228, #232 (amendments), #234, #235.

Refs #560

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@tae898
tae898 merged commit a04fb1d into main Nov 3, 2026
4 checks passed
@tae898
tae898 deleted the docs/method-2611-amendments branch November 3, 2026 15:23
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Method docs: 2026-11-03 amendments (idle hold 0 us window, excluded windows, run order, paper dates)

1 participant