The root handoff, rewritten for the build — and the walkthrough archived rather than deleted - #477
Merged
Merged
Conversation
…o it moves rather than being deleted `pipeline.md` landed in #476 and supersedes the walkthrough's plan, so the root handoff had to change. It is archived rather than deleted: its field-step section and its trap list are still the best account of both, and the third option exists precisely for this — a file spent as a signal and live as a record. That option was missing until 2026-08-24 and its absence cost four days of a false positive. The new one names the same PR, #466, because that is still the only genuinely open piece of work in the tree and its closure is still the honest retirement condition. **It leads with the summary page, because Tony has asked for it twice, it is still not right, and nothing blocks it.** What is wrong is the geometry — FAST stacked over SLOW at full width, so a page runs to nine thousand pixels and cannot be read at once. interface2 solved that in July and extracted the renderer so a second caller would not have to redraw it. The handoff points at the contract to port and names the three things the current version drops that carry information: the two shading levels per region, region names as text, and the counts table. It carries the colour change explicitly, because the old figure will mislead anyone who opens it: red there means an event in no coordinated event, with colour otherwise encoding width. Going forward the raster is black and red means excluded. **Two blockers are stated once, at the top, rather than discovered per step.** Nothing persists a trained model, which blocks testing a variant on a fresh batch, running one on the user's folder, and accepting a model the user brings. And the library's detect path has no settings argument, which is where the two modes stop being one pathway — the browser can apply a tuned operating point to a folder and the command line cannot. The traps are the ones that actually cost time in the thread that produced it, including two of my own: inventing step names for steps the webapp's rail already named, and guessing a repo from its directory name while hunting an estimator that was in `assess.py`. Co-Authored-By: Claude Opus 5 (1M context) <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.
docs/pipeline.mdlanded in #476 and supersedes the walkthrough handoff's plan, so the root signal had to change.The old one is archived, not deleted. Its field-step section and its trap list are still the best account of both. That is the third option
docs/handoffs/README.mdoffers — a file spent as a signal and live as a record — added on 2026-08-24 after its absence cost four days of a false positive.The new file names the same PR, #466, because that is still the only genuinely open piece of work in the tree and its closure is still the honest retirement condition.
What it leads with
The summary page. Tony has asked for it twice, it is still not right, and nothing blocks it. The defect is the geometry —
FASTstacked overSLOWat full page width, so a page runs to nine thousand pixels and cannot be read at once. interface2 solved this in July and extractedrender_coord_summary_page.mso a second caller would not redraw it; the handoff points at the contract to port and names the three things the current version drops that carry information: two shading levels per region, region names as text, and the counts table.It also states the colour change explicitly, because the MATLAB page will mislead anyone who opens it — red there means an event belonging to no coordinated event, with colour otherwise encoding width. Going forward the raster is black and red means excluded.
Two blockers, stated once instead of discovered per step
Nothing persists a trained model — blocks testing a variant on a fresh batch, running one on the user's folder, and accepting a model the user brings.
The library's detect path has no settings argument — the browser can apply a tuned operating point to a folder and the command line cannot. That is where the two modes stop being one pathway.
Traps
The ones that actually cost time in the thread that produced this, including two of mine: inventing step names for steps the webapp's rail already named, and guessing a repo from its directory name while hunting an estimator that was in
assess.py.Sapper clear; handoff-honesty and index-resolution gates green.
🤖 Generated with Claude Code