Eleven goals chartered for the 95% sprint; full-WPT pass rate reaches 80.57%
Eleven unattended goals were chartered in two phases toward 90%/95%; wptfyi-webcrypto and wptfyi-runtime hit their gates and archived within three days, joining three earlier goals for five archives on the week. The full-WPT pass rate rose from 68.36% to 80.57%. The rendering corpus grew to 17,049 and its failures 691→995 track that growth; after the R5603 knives the A/B reads FIXED=1 REGRESSED=0.
Overview
The week brought 930 commits, five goals archived, and the wptfyi sprint charter. On 11-09 eleven unattended goals were chartered in one pass: 01–05 race to 90% (CSS +153,000; Editing/Selection/Input +130,000; HTML +81,000; runtime +76,000; WebCrypto +31,000), 06–11 race to 95% (DOM +35,000; network +33,000; data +25,000; graphics +22,000; automation +10,000; media +10,000), with the 11-07 snapshot (68.36%) as the shared baseline and acceptance basis.
Two sprint goals hit their gates and archived within three days: wptfyi-webcrypto (11-10) netted +41,098 across three milestones; wptfyi-runtime (11-11) put the test262 full baseline at 53,166 cases (96.39%), the DC-3 union of three directories at 1,393,984 against the 1,393,955 acceptance line, net +76,029 over 11-07, and attributed its 7 in-ledger failures one by one.
Three goals chartered earlier also archived: shadow-dom-rendering no. 45 (11-08, M0–M4 done, directory moved to archive/), http-cache no. 47 (11-09, the archive re-run read three accounts 106P/0F · 286P/0F · 124P/14F with zero drift), runner-net-infra no. 46 (11-09, wss batch gate 489/489 = 100%, all four DCs passed).
The full-WPT accounting rose to 80.57% in the 11-14 weekly snapshot (1,830,410/2,271,889 subtests), from 68.36% (1,551,067/2,268,984) a week earlier — roughly 279,000 net subtests; covered files 71,480 (56.24%), all-pass files 61,688 (48.54%).
The 11-08 weekly candidate snapshot was intercepted by the aggregation guard (custom-elements and shadow-dom passing dropped to zero) and stayed unrecorded; the 11-12 and 11-14 snapshots published normally. wptfyi-css carries the sprint's largest gap (166,939): DC-1 closed with the gap split into three buckets — about 139.5k in-face assertion failures, about 18.6k un-wired corpus, plus an out-of-channel remainder; DC-2 landed 13 corpus directories at missing=0; the snapshot moved from 52,488 (26.70%) on 11-10 to 62,896/197,163 (31.90%) on 11-14, net +10,408 in four days.
The 11-15 target expando knife unlocked the interpolation harness domain-wide: group-by-group measurement across 337 pages unlocked 335 groups, still-declared-0 cleared to zero, and the unlocked face reads 6,691P (css-masking 126P→615P with zero page regressions); g41 gradient canonical +552, g45 image-set cleared its 114 failures, g47 closed the last 21 css-images failures, g49 GCS viewport +72 (css-values 4,185P); the anchor-position corpus batch brought first coverage of 12,061P/792F. wptfyi-editing was chartered on 11-09; DC-1 settled and the weekly snapshot reads 109,059/154,595, net +92,875 over the 11-07 baseline. editing-compat kept cutting the insertParagraph deep water, 99F→16F, the latest ip C6+C3 non-text atomic-split boundary gate batch all five legs green.
The rendering corpus grew to 17,049 (anchor-position batches); R5603 accounting reads 995F with FIXED=1 REGRESSED=∅ — the 691→995 rise tracks the denominator, not regressions; the flex-item va real regression introduced by the g50 knife was bisected to 7bc31687d and fixed in-round.
The ROADMAP rendering line moved to the R5471 accounting (15,911/16,599 ≈ 95.8%, fail 688). html-semantics-compat moved to a v2.0 control plane (M2 slice 455 closed); hold patrols s433–s457 landed nothing remote, the P set has held at 8,306 for four consecutive rounds, and the all-bridge green window held 1,643P/0F.
Completed
wptfyi sprint chartered 11 unattended goals on 11-09: 01–05 to 90% (CSS +153,000, Editing +130,000, HTML +81,000, runtime +76,000, WebCrypto +31,000); 06–11 to 95% (DOM +35,000, network +33,000, data +25,000, graphics +22,000, automation +10,000, media +10,000)
wptfyi-webcrypto archived on 11-10 with three milestones netting +41,098 against the +31,000 gate; wptfyi-runtime archived on 11-11 with test262 at 53,166 cases (96.39%), DC-3 union 1,393,984 ≥ acceptance line, net +76,029 against the +76,000 gate, 7 in-ledger failures attributed one by one
shadow-dom-rendering no. 45 moved to archive (11-08); http-cache no. 47 archive re-run read 106P/0F · 286P/0F · 124P/14F with zero drift (11-09); runner-net-infra no. 46 closed with the wss batch gate 489/489 = 100% and all four DCs passed (11-09)
Full-WPT accounting 68.36% (11-07) → 80.57% (11-14, 1,830,410/2,271,889 subtests, covered files 71,480 at 56.24%, all-pass 61,688 at 48.54%); the 11-08 candidate snapshot was intercepted by the aggregation guard (custom-elements/shadow-dom passing dropped to zero) and stayed unrecorded
wptfyi-css DC-1 closed (gap in three buckets: ~139.5k in-face assertion failures, ~18.6k un-wired corpus) and DC-2 landed 13 corpus directories at missing=0; snapshot 52,488 (26.70%, 11-10) → 62,896/197,163 (31.90%, 11-14), net +10,408 in four days; anchor-position corpus batch first coverage 12,061P/792F
target expando knife unlocked the interpolation harness domain-wide (335/337 groups, still-declared-0 cleared, unlocked face 6,691P, css-masking 126P→615P with zero page regressions); g41 +552, g45 image-set 114 failures cleared (+143), g47 closed the last 21 css-images failures, g49 GCS viewport +72 (css-values 4,185P)
wptfyi-editing chartered 11-09, DC-1 settled, weekly snapshot 109,059/154,595 for net +92,875 over the 11-07 baseline; editing-compat cut the insertParagraph residual 99F→16F (ip C6+C3 gate batch, five legs green, F 20→16)
Rendering corpus 16,594→17,049 (anchor-position batches); R5603 accounting 995F with FIXED=1 REGRESSED=0, the flex-item va real regression from the g50 knife bisected to 7bc31687d and fixed in-round; ROADMAP rendering line refreshed to R5471 (15,911/16,599 ≈ 95.8%); html-semantics-compat control plane v2.0 with patrols s433–s457 (P set held at 8,306 for four consecutive rounds, all-bridge green window 1,643P/0F)
In progress
wptfyi-css's next knife is the WAAPI gate (21.7% of the 50,731 remaining failures), followed by the supports-coverage family (49.5%) and interpolation semantics (28.6%), attribute family by A/B; the sprint's other nine goals (editing, html, dom, network, data, graphics, automation, media among them) are running; rendering-compat has 995 cases left to 100% (denominator 17,049); four idle three-source check rounds have followed R5603 with the corpus skipped, and the next accounting waits for the peer side to land again for a full corpus A/B.
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.