Editing sprint goal archived, CSS face passes halfway; full-WPT pass rate reaches 84.08%
wptfyi-editing archived with net +130,043 past the +130,000 budget and editing-compat closed the same week; the css weekly audit reads 101,677/199,018 (51.09%) for net +72,285, and html's the-xhtml-syntax channel went from zero-accounted to 25,606/25,606. The full-WPT pass rate rose from 80.57% to 84.08%. The rendering baseline r5603 (995F) held all week.
Overview
The week brought 1,253 commits.
Sprint goal 2, editing, crossed the line: wptfyi-editing archived on 11-20 with the DC-3 final audit reading 146,227 against the acceptance line of 146,184 (margin +43), net +130,043 over the 11-07 baseline, clearing the +130,000 budget with no attributable regressions nearby; editing-compat archived a day earlier — its R24 DC-5 coverage re-test lifted editing topdir coverage from 11.41% to 92.5%, editing/run file coverage from 0/120 to 119/127, and the three-state unaccounted count to zero, closing at 117,244/108,448.
Goal 1, css, passed the halfway mark: the weekly DC-3 audit reads 101,677/199,018 (51.09%), net +72,285 over the 29,392 baseline against the +153,000 target (about 80,700 to go).
Knives kept accounting: the CSSOM-INSETS-USEDECHO knife landed the gCS inset resolved=used read-back chain (absolute/fixed pages +96 flip), the fontcycle ffix pieces closed at net +154 on the canonical audit, and the hyphens M4-a legs closed their face while M4-b's one 16-P2F reading was re-judged flake on double-witness evidence (tally unchanged).
Goal 3, html, finished DC-2's fourteen batches: DC-1 first took the the-xhtml-syntax entity-matrix channel from zero-accounted to 25,606/25,606 with zero regressions, and batches one through thirteen advanced cluster by cluster (html/dom +1,888, rendering +2,551, webappapis +1,537, semantics +1,955, canvas/offscreen +1,062, infrastructure +623 among others); the 11-21 snapshot reads the html face at 102,884/160,118 (64.26%), net +31,788 over the 71,096 baseline against the +81,000 target.
The full-WPT accounting rose to 84.08% in the 11-21 weekly snapshot (1,913,571/2,276,027 subtests), from 80.57% a week earlier; covered files 73,282 (57.50%), all-pass files 63,390 (49.74%).
Rendering held its ground all week: the r5603 baseline (995F, corpus 17,049) stayed in place; the R5745 ignition round re-audited the three knives that had slipped past pull-window accounting (the css-gaps animation channel plus two ilb WSCANON commits) and read stable 995 SET-IDENTICAL, with the hold batches carrying that through R5844; the ROADMAP rendering line moved to the 11-15 accounting (upstream corpus default 15,914/16,594 ≈ 95.9%).
Engine-side the week saw 115 fix(engine), 28 fix(editing), 18 fix(style-system) and 17 feat(engine) commits — the css and html face gains rest on those fixes. html-semantics-compat's hold patrol reached round s621 and closed M2 slice 620 (join stable 8,517: 8,305P/206F/5T/1U with fully zero displacement); the all-bridge green window held 1,660P/0F. Sprint goals 06–11 (dom, network, data, graphics, automation, media) remain in preparation ahead of their first full-round confirmation.
Completed
wptfyi-editing archived on 11-20: DC-3 final audit 146,227 ≥ acceptance line 146,184 (margin +43), net +130,043 over the 11-07 baseline against the +130,000 budget, no attributable regressions nearby
editing-compat archived on 11-19: R24 DC-5 coverage re-test closed — topdir 11.41%→92.5%, editing/run 0/120→119/127, three-state unaccounted=0, closing measurement 117,244/108,448
wptfyi-css weekly DC-3 audit at 101,677/199,018 (51.09%), net +72,285 over the 29,392 baseline (target +153,000); CSSOM-INSETS-USEDECHO knife flipped absolute/fixed pages +96, fontcycle ffix closed at net +154 on canonical audit
wptfyi-html DC-1 took the the-xhtml-syntax entity-matrix channel 0→25,606/25,606 with zero regressions; DC-2 batches one through fourteen done, batches one to thirteen advancing cluster by cluster (+1,888/+2,551/+1,537/+1,955/+1,062/+623 among others); 11-21 snapshot reads the html face at 102,884/160,118 (64.26%)
Full-WPT accounting 80.57% (11-14) → 84.08% (11-21, 1,913,571/2,276,027 subtests, covered files 73,282 at 57.50%, all-pass 63,390 at 49.74%)
Rendering held all week: r5603 baseline (995F, corpus 17,049) in place; R5745 re-audited the three knives that slipped past window accounting and read stable 995 SET-IDENTICAL, hold batches carrying through R5844; ROADMAP rendering line moved to the 11-15 accounting (15,914/16,594 ≈ 95.9%)
html-semantics-compat patrol reached s621 and closed M2 slice 620 (join stable 8,517 = 8,305P/206F/5T/1U with fully zero displacement); all-bridge green window held 1,660P/0F
Engine side: 115 fix(engine), 28 fix(editing), 18 fix(style-system) and 17 feat(engine) commits, carrying the css/html face gains
In progress
wptfyi-css has about 80,700 left to the +153,000 target and works down the residual buckets (the CSSOM-INSETS v2 pool with 336 failures across five root causes, plus the interpolation-semantics and supports-coverage knife families); wptfyi-html prepares for the DC-3 full audit; goals 06–11 (dom, network, data, graphics, automation, media) await their first full-round confirmation; rendering-compat's hold batches continue (R5844's next step is the kb85 deployment launch).
Sources
Engineering guides
How ZeroWeb tests AI-autonomous software engineering
A browser gives AI development a long-running test case: architecture, code, debugging and verification must work together.
ZeroWeb Rust browser architecture: from request to pixels
Follow the page pipeline and the boundaries between the browser application, embeddable WebView, engine and rendering infrastructure.
Getting started with ZeroWebView and ZeroBrowser
Prepare a local checkout, launch the desktop browser and inspect the WebView API before embedding the experimental engine.
How to read ZeroWeb WPT and performance results
Understand self-source reftests, assertion-based suites, changing corpus sizes and p95 benchmarks before comparing results.