The fair comparison's four architectures, drawn through draughtsman at one slot - #660
Merged
Merged
Conversation
…t one slot WSMIP064 and WSMIP065 are each writing a report on goal 2's fair comparison, and both briefs asked for a figure of what the four architectures do differently. Tony: "They should definitely use draughtsman otherwise they'll likely be completely incomparable." So chorus_norm, chorus_gain_norm and line_length get specs beside a comparison sibling of the tube's, all at report.css's 912px column with a 9.5px type floor, and every one breaks its rows at the same place: the first row is the stages that still hold one row per ROI, the second starts where that axis collapses, the third is the head. The first row's length is the comparison: two boxes for tube, three for line_length, four for both chorus. TRACED THROUGH THE REGISTRY. build_chorus_norm() called bare returns plain chorus — the registered norm=True never reaches it — and draughtsman calls its target with no arguments, so the first trace was of the failed control and check passed on it. bugarach.learn.registered exposes ARCHITECTURES[name].make as a zero-argument target. tube and line_length are safe only because their builders' defaults equal their registered config. Named row breaks needed a draughtsman change (layout.breaks, draughtsman PR #1 at 18b4453); re-vendored here. The front-page specs change only their stamp, which the freshness family requires to agree, and architecture.svg and architecture-phone.svg are byte-identical. tools/make_comparison_figure.py stacks the four committed figures at their own extents, one unit to one pixel, and refuses a set whose specs disagree about the slot. It writes to the darkroom; the page is not committed, the four figures it is built from are, and the staleness test regenerates them. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
WSMIP064's fair-comparison report ships in the public repo and points at the architecture drawings by path, which works for the darkroom page only on a machine with the mount. docs/learned/comparison/comparison.svg is the four committed panels composed at one scale, rewritten by --commit and byte-compared against a fresh composition by the test, so it cannot outlive its panels. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This was referenced Sep 19, 2026
Merged
syncytium2
pushed a commit
that referenced
this pull request
Sep 19, 2026
…prescribed was wrong The base is green. WSMIP064 marked the tuning module serial in b9752a9 and run 1624 passed; #660 then passed on the repaired base and merged, which is why the branch has grown to 66 commits and 475 files since the scan was written. The gate told the repair to drop the inner --jobs to 1 as well. 064 measured that before doing it: the quick run takes 276 s at four jobs and 559 s at one, and 559 against a 600 s budget leaves no room on a slower runner. The jobs stayed at four. Recorded here because the scan is what a later session will read. The consequence is worth naming: a serial module runs after the parallel pass with the runner to itself, so any branch carrying this fixture now runs a two-hour suite instead of ten minutes. That cost arrives on main with this branch. Gates 2 through 5 are unchanged and open. The username is still at line 682 and five handoff files still sit at the root, so the closes-when list now says which are done. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017hQX6iESxBsQk7Jan875e3
This was referenced Sep 19, 2026
syncytium2
added a commit
that referenced
this pull request
Sep 19, 2026
#662) * The landing scan for tune-bench-comparison: five gates, and nine tenths of it is run output Tony asked what has to be true before the weekend's branch can reach main. Every figure is measured against main at ab4eef1 rather than estimated. The headline is that the branch is not dirty. It merges with no conflict, sapper reports nothing and check_quotes is clear. And "54,000 lines" is misleading: the tool, its tests and the library changes are 3,742 lines across 21 files, while 429 files and 49,509 insertions are what the runs produced. Five gates, of which only one is a repair. The base is red because a module-scoped fixture shells out with a 600-second budget and four jobs, unmarked serial, under a four-worker parallel run on a four-core runner — so the budget measures the other workers, which is the shape CLAUDE.md already has a rule for. One is a ruling Tony reserved and would otherwise inherit by accident: this branch has #596 as an ancestor, and #596 asks to be held if the new models should stay out of the picker. One is a personal path in a public repo that SAP004 cannot see. One is the root handoffs. The last gate is the question worth asking first: the seven stranded architectures are 7 files of the 454. If un-stranding them is the goal, a nets-only branch does it today and inherits neither the reserved ruling nor the run output. Landing everything is a larger and separate act. * Re-measure the landing scan: gate 1 is repaired, and half of what it prescribed was wrong The base is green. WSMIP064 marked the tuning module serial in b9752a9 and run 1624 passed; #660 then passed on the repaired base and merged, which is why the branch has grown to 66 commits and 475 files since the scan was written. The gate told the repair to drop the inner --jobs to 1 as well. 064 measured that before doing it: the quick run takes 276 s at four jobs and 559 s at one, and 559 against a 600 s budget leaves no room on a slower runner. The jobs stayed at four. Recorded here because the scan is what a later session will read. The consequence is worth naming: a serial module runs after the parallel pass with the runner to itself, so any branch carrying this fixture now runs a two-hour suite instead of ten minutes. That cost arrives on main with this branch. Gates 2 through 5 are unchanged and open. The username is still at line 682 and five handoff files still sit at the root, so the closes-when list now says which are done. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017hQX6iESxBsQk7Jan875e3 --------- Co-authored-by: Claude <noreply@anthropic.com>
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.
Specs and
DRAWABLEentries forchorus_norm,chorus_gain_normandline_length, plus a comparison sibling of the tube's spec. All four are at report.css's 912px column with a 9.5px floor, and each breaks its rows where its ROI axis collapses. The PR targetstune-bench-comparisonbecause the two chorus nets exist only there.build_chorus_norm()called bare builds plainchorus. The newbugarach.learn.registered:<name>target traces what the registry trains.18b4453(row breaks, A spec names where a wrapped figure starts a new row draughtsman#1). The front-page specs change only their stamp, and both front-page SVGs are byte-identical.tools/make_comparison_figure.pybuilds the one-scale page into the darkroom, under claim Claim a darkroom folder for the four comparison architectures, before drawing them #658.tests/test_comparison_figure.pycovers the registry target, the single scale, the refusal on mismatched slots, and named breaks. The staleness test picks up the new figures fromDRAWABLE.Locally, with the worktree's own
srcfirst on the path, the four relevant test files pass (44). A plain worktree run fails 24 tests. 12 of those also fail on cleantune-bench-comparison(13 fail there), from the known issue that worktrees import the primary checkout'ssrc. Two more are this PR's own registry tests failing for the same reason, and they pass with the worktree'ssrc. The other 10 were not re-run against the baseline. CI is the verdict.🤖 Generated with Claude Code