Repository navigation
Share and restore Explore experiment settings - #251
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
Dylan Roy — delegated AI review. Actual performer: OpenAI/Codex /root/explore_state_review, fresh nonauthor source review. PASS: no actionable engineering findings on base61395649914a98f8aec40ee151e94c34983f5015, headfcee4737b959e2348e7a894e866e4a9cea25608b, tree18a0a48f06ed2e25849a21b0057ff4da522c9db1. The reviewer independently fetched live GitHub source/diff identities. All22 staged changed-file blobs match GitHub; normalizing only generated site-release keys leaves exactly five substantive files. Model arithmetic, equations, scientific sources and status remain unchanged. Independent reviewer checks:178 Node passes and2 explicit pre-existing skips,21 Python passes, asset-release consistency. Parser/serialization covered all138 numeric control positions and8 object subsets, plus10 additional malformed inputs. Execution of the actual explore.mjs in an in-memory DOM/event harness passed initial restoration, input/change commits, Back/Forward event restoration, fragment refresh/preservation, resets, malformed fallback and refused-history share-link fallback. Source inspection confirmed invalid-field notices, preserved valid fields, disabled static controls, hidden fallback share link and native keyboard-accessible anchor. Reviewer exposure limit: the event harness does not establish native browser event ordering, visual behavior or assistive-technology behavior. The hosted browser test preserves an initial fragment but does not navigate the Explore lesson fragments; that native path remains a separate live check. The reviewer did not execute or adopt the author's hosted-browser results as its own. No reviewer edits. Same-provider organizational-independence credit0; human reading PENDING. Scientific effect NONE. Separate author/hosted evidence: full applicable Node run178PASS/2existingSKIP,26PythonPASS; four new URL tests failed before implementation. All8 PR-head hosted checks PASS. Browser run37146171976 passed16/16 Chrome154.0.8037.57 cases, actual PR merge9a21cd938e4b95073aa9a7537a9b4265aaaec7c1/tree18a0a48f06ed2e25849a21b0057ff4da522c9db1. Artifact11282835222 downloaded exactly9893409B SHA2562e98cdb337a8deabb8c1fb146484baf2a931681e045b1518ab7a4f16f8bdef54. Four Explore desktop/mobile light/dark renders inspected; no horizontal overflow. This evidence does not substitute for deployed Pages readback. |
|
Completed and RELEASED — native Explore-state pickup #250 (comment). PR251 merged with expected head fcee473 as a414e77. Fresh landed tree18a0a48f06ed2e25849a21b0057ff4da522c9db1 equals the reviewed candidate and hosted PR merge tree; actual parents61395649914a98f8aec40ee151e94c34983f5015 andfcee4737b959e2348e7a894e866e4a9cea25608b. Five substantive files and17 generated release-key-only files. Existing teaching arithmetic, equation/source text, inventories and scientific status unchanged. Full scoped nonauthor engineering PASS: #251 (comment). Actual reviewer OpenAI/Codex /root/explore_state_review; independent source and in-memory event checks, no native-browser claim. Same-provider independence credit0; personal human reading PENDING. Automatic PR review5972446910 separately reports completed at19:00:34Z with no posted inline findings. Verification: full applicable Node suite178PASS/2existingSKIP; author26PythonPASS; new URL regressions failed before implementation. Hosted PR browser run37146171976 passed16/16 Chrome154.0.8037.57 cases. Exact downloaded artifact11282835222:9893409B SHA2562e98cdb337a8deabb8c1fb146484baf2a931681e045b1518ab7a4f16f8bdef54. Four Explore renders inspected at1200/390 widths, light/dark. All8 PR-head checks passed. Fresh all8 landed check-runs at a414e77 passed, including public-shop/browser-smoke37146353698, landing/formal/verify37146353880 and Pages37146353397. Separate deployed-browser readback: entered Explore from public navigation, used ten keyboard steps to s=-1, navigated to the distance lesson, selected r=0.25 and remote, navigated to the finite-object lesson and removed Object2, then reopened the actual share link. Values restored: s=-1, R=1, r=0.25, remote, objects1/3; eigenvalue boundary is flat, gap0.015625, two valid groups. Actual lesson-fragment Back/Forward preserved those states. Verified public example: Local staged source /workspace/scratch/5107d3a6aa11/explore-state; hosted evidence /workspace/scratch/5107d3a6aa11/explore-browser. Scope released. No C114/R17 ownership inferred, no archive duplication or source promotion. Remaining UI roadmap and retained Library/equation owner continue separately. Scientific effect NONE. |
|
Pickup — citation source-reference usability, actual executor OpenAI/Codex /root, continuity PR246/247/248/250/251. Base a414e77. Fresh live open PRs and native comments checked: PR249 workflow integration belongs to C118; retained Library/equation-reader scope remains separate. Bounded native lease through 2026-10-03T21:10:00Z. Paths: docs/site/{cite.html,cite.mjs,source-reference.mjs,reference.css}, tests/test_source_reference.mjs, tools/public_shop_browser_check.py; generated site-release substitutions only in other already generated files. Add optional supplied SHA-256, copyable plain/JSON source identity, explicitly unverified user-entered metadata, share/reload/Back/Forward for built references, invalid-input and clipboard fallback. No invented authorship, scientific/review status, fetched bytes or bibliography completeness. No Library/source/proof/status changes. Isolated branch codex/citation-reference-20261003; substantive fresh review and current hosted/browser checks before guarded integration. This does not claim R17/C114, PR228 writer authority or a Drive archive. |
Explore links now retain the existing curvature, distance/region and finite-object controls. Valid links restore controls and derived explanations; malformed or duplicate fields restore starting values with a visible notice. Keyboard changes, resets and Back/Forward keep controls and URLs in agreement. JavaScript-off readers see an explicit static-example boundary.
Scope pickup: #250 (comment). Actual author OpenAI/Codex /root, separate from retained Library/equation scope. Five substantive files; remaining changes are generated asset-release keys. No arithmetic, source, equation, research-status or inventory changes. Scientific effect NONE.
Validation: 80 Node and26 Python tests PASS; four new URL tests failed before implementation. Hosted browser controls cover all three models, empty selection, keyboard edits, reset, reload, history, malformed fields and JavaScript-off fallback in existing desktop/mobile light/dark cases. Exact-head nonauthor review and current hosted checks are required before guarded integration.