Html api docs improvement#61

Draft

sirreal wants to merge 193 commits into

trunkfrom

html-api-docs-improvement

Commits on Jun 11, 2026

HTML API docs experiment: plan contract and markdown renderer.
sirreal
committed
HTML API docs experiment: task corpus and execution harness.
sirreal
committed
HTML API docs experiment: round tooling and protocol.
sirreal
committed
HTML API docs experiment: workflow scripts and trial persistence.
sirreal
committed
HTML API docs experiment: round 0 baseline results.
sirreal
committed
HTML API docs round 1, hypothesis 1: closer-token depth semantics.
sirreal
committed
HTML API docs round 1, hypothesis 2: rehabilitate HTML Processor next_token().
sirreal
committed
HTML API docs round 1, hypothesis 3: get_modifiable_text() returns decoded text.
sirreal
committed
HTML API docs experiment: round 1 results — all hypotheses confirmed.
sirreal
committed
HTML API docs experiment: corpus revision per review.
sirreal
committed
HTML API docs round 3, hypothesis 1: set_attribute() placement rules.
sirreal
committed
HTML API docs round 3, hypothesis 2: correct the class-level support claims.
sirreal
committed
HTML API docs round 3, hypothesis 3: serialize_token() rewrite idiom.
sirreal
committed
HTML API docs experiment: round 2 Haiku baseline results.
sirreal
committed
HTML API docs experiment: ingest tooling for the long-run loop.
sirreal
committed
HTML API docs round 4, hypothesis 1: serialization is not how you read edits.
sirreal
committed
HTML API docs round 4, hypothesis 2: which tokens carry modifiable text.
sirreal
committed
HTML API docs round 4, hypothesis 3: bookmark name reuse is the last-X idiom.
sirreal
committed
HTML API docs experiment: round 3 results — refine serialize guidance, two new gaps.
sirreal
committed
HTML API docs round 5, hypothesis 1: building markup from a template.
sirreal
committed
HTML API docs round 5, hypotheses 2-3: next_tag() matching contract; decoded reads; add_class idempotency.
sirreal
committed
HTML API docs round 5, hypothesis 4: why the subtree walk uses >=.
sirreal
committed
HTML API docs experiment: round 4 results — train 94.18 (+3.5), T07 cured.
sirreal
committed
HTML API docs round 6 hypotheses: processor chooser, tree-awareness boundary, get_updated_html identity, >= warning placement.
sirreal
committed
HTML API docs experiment: round 5 results — train 94.77, T04 +49.2.
sirreal
committed
HTML API docs round 7 hypotheses: RCDATA text location on the HTML Processor, >= beside the operator, drain idiom, add_class return semantics.
sirreal
committed
HTML API docs experiment: round 6 checkpoint — train 97.84, held-out above baseline.
sirreal
committed
HTML API docs round 8 hypotheses: UTF-8 output statement; break-form boundary rule.
sirreal
committed
HTML API docs experiment: round 7 results — train 97.51, N03 perfect.
sirreal
committed
HTML API docs round 9 hypotheses: the shared cursor, and surfacing the last-X bookmark idiom.
sirreal
committed
HTML API docs round 9, hypothesis 2: surface the last-X bookmark idiom in the bookmarks narrative.
sirreal
committed
HTML API docs experiment: round 8 results — train 97.70, first saturated task.
sirreal
committed
HTML API docs round 10 hypothesis: the RCDATA exception belongs on the walk path.
sirreal
committed
HTML API docs experiment: round 9 checkpoint — train 98.66, T08 stabilized.
sirreal
committed
HTML API docs round 11 hypotheses: the equality case is the reason for >=; empty regions flush naturally.
sirreal
committed
HTML API docs experiment: round 10 results — train 98.70, T08 perfect.
sirreal
committed
HTML API docs round 12 hypotheses: add_class never removes; quoting scope co-located.
sirreal
committed
HTML API docs experiment: round 11 results — train 98.28, T09 perfect.
sirreal
committed
HTML API docs round 13 hypotheses: no closer-guard needed by default; implied elements appear in walks.
sirreal
committed
HTML API docs experiment: round 12 checkpoint — held-out 91.04, N05 +12.4.
sirreal
committed
HTML API docs round 14 hypothesis: post-collection measurement noted in the recipe.
sirreal
committed
HTML API docs experiment: round 13 results — first 100% functional sweep, train 98.54.
sirreal
committed
HTML API docs round 15 hypothesis: construction asymmetry stated on both classes.
sirreal
committed
HTML API docs experiment: round 14 results — construction-asymmetry gap now train-evidenced.
sirreal
committed

Commits on Jun 12, 2026

Commits on Jun 13, 2026