v3.2 verification
Status records observed evidence, not implementation intent. The date reference uses the academic homepage news date; the linked blog Archive uses a different date format.
| Item | Status | Evidence |
|---|---|---|
| P0-1 Persistent weighted question deck | Verified in tests and Edge; refresh policy not verified | tests/test_intro_deck.cjs (1,000 seeds, 72 visits; 300-visit ratio); output/playwright/v3.2/deck-browser.json (6 visits) |
| P0-2 Typing and real typo corrections | Verified in tests and Edge at 375px | tests/test_intro_typing.cjs; output/playwright/v3.2/typing-browser.json and typing-forced-*.png; WebKit not verified |
| P0-3 Figure flow, anchors, states, exit, renderer | Verified in Edge; WebKit and touch not verified | tests/test_figure_flow.py; output/playwright/v3.2/figure-{nojs,snap,exit,return}-check.txt; desktop and mobile screenshots. Mobile calibration places two inputs side by side before a downward merge. |
| P0-4 Recent Updates date style | Verified against reconstructed reference, not direct live browser | docs/refs/edwinos-date.md; output/playwright/date-*-{1440,375}-{light,dark}.png; tools/test_archive_markdown.py |
| P1-1 Intro success icon | Verified in Edge | output/playwright/v3.2/intro-p1-sequence.gif, intro-p1-timings.json; parameters tuned by eye in docs/refs/apple-pay-success.md |
| P1-2 Grid ripple and cell morph | Verified in Edge | output/playwright/v3.2/intro-p1-sequence.gif |
| P1-3 Transition candidate comparison | Verified in Edge | output/playwright/v3.2/transitions/dock.gif, shutter.gif, debug-search.json; shutter chosen because dock compresses the question and travels upward |
| P1-4 Figure skill update | Verified | .agents/skills/figure-to-animation/scripts/test_validate_spec.py (4 tests); hooks/figure_specs.py checks production specs at build |
| P2 Responsive browser matrix and recording | Partial: Edge verified; WebKit not verified | output/playwright/v3.2/p2-edge-{375,768,1440}.png, p2-edge-scroll.webm. WebKit executable missing; download stalled. |
| P2 Reduced motion, LCP and CLS | Partial | Edge p2-edge-reduced.png and p2-metrics.json; no before baseline, so regression is not verified. |
| P2 Hero first viewport | Verified in Edge at 1440×900 | output/playwright/v3.2/hero-1440.png; video and clip controls fit the viewport |
| P2 Reference notes | Partially verified | docs/refs/ contains notes; reference-site scroll recordings and runtime measurements from v3.1 remain incomplete |
Detailed scope and limitations
- P0-1: fixed first question removed; normal entries include sim-gap; localStorage deck, ID hash rebuild, 2:1 group weights and category rotation tested.
qandseeddebug handling:output/playwright/v3.2/transitions/debug-search.jsonandtests/test_intro_deck.cjs. Cross-browser storage behavior and ordinary refresh policy remain not verified. - P0-2: typo probability and weighted types, eligible characters, correction lag, timing cap, final text and seed exercised in
tests/test_intro_typing.cjs. Edge 375px typing, no-typo and reduced-motion samples:output/playwright/v3.2/typing-browser.jsonandtyping-reduced.png. Layout stability at other widths and WebKit remain not verified. - P0-3 A–H: monotone FigureSpec validation, measured ResizeObserver anchors, opaque static-position nodes, connector flow, step return, pin release, no-JS and Edge responsive geometry tested in
tests/test_figure_flow.pyandoutput/playwright/v3.2/figure-{nojs,snap,exit,return}-check.txt. Mobile calibration retains two parallel inputs instead of a strict single column. Continuous touch, WebKit, all animation timing and exact connector anchors at every intermediate resize remain not verified. - P0-4:
docs/refs/edwinos-date.mddescribes live HTML/CSS reconstructed in Playwright with local fonts and measured light/dark styles. Direct external browser loading timed out. The academic news reference is MM/YYYY; its linked Archive uses a different format. Mobile blog rows stack intentionally. - P1-1/2: icon, ripple and cell morph appear in
intro-p1-sequence.gif; Apple timing values are visual tuning, not measurements. Reduced-motion and skip were checked in Edge. - P1-3: the dock and shutter comparison GIFs are screenshot sequences, not continuous recordings; shutter was selected.
- P1-4: the generic skill retains the license gate; tests cover invalid coordinates and backwards edges.
- P2: a current-versus-before LCP/CLS comparison has not been performed; isolated lab measurements alone cannot prove no regression.
The recorded full-page scroll reached the footer, but the external visitor badge image had not finished loading in its final frame; badge availability was not established by that recording.