- Add configurable PopClip-style selection bar (Explain this, Ask, Check me, Taxonomy, Simplify, Example, Study tools) with gear menu and Learning preferences sync
- Add security-hardened BYOK and 1-Click OpenRouter OAuth PKCE SSO module (src_shared/byok-provider.js) with strict HTTPS host allowlist, secret redaction, session-only default storage, and shared .github.io origin protection
- Add Free shared model status bar and Connect your own Gemini / Claude controls in both Focused Reader and Study Tools
- Extract interactive figure explorer ([data-mlsys-explore]) control parameters in reader and study context extractors
- Update SocratiQ user guide prose and 2x HiDPI screenshots and add unit and Playwright regression coverage
- Add socratiq/scripts/sync-taxonomy.mjs and generated socratiq/src_shared/taxonomy-map.js from taxonomy/framework-data.json
- Ground Focused Reader and Study Tools system prompts in the 5-Group ML Systems Taxonomy (F, S, P, Q, T) and active chapter subjects across Volumes I-IV
- Add 'Map to taxonomy' reader action, Taxonomy filter in + Context and @ mentions, and F/S/P/Q/T Knowledge Graph taxonomy alignment
- Harden responsive layouts across 320x568-1440x900 viewports (Light & Dark modes) including #tbl-socratiq-controls mobile wrapping, modal headers, and drawer sizing
- Add automated regression tests for chapter taxonomy resolution, reader taxonomy action, and + Context / @taxonomy mentions
- Rename navigation entry to 'AI Companion (SocratiQ)' across Volumes I-IV (_quarto-html-vol{1..4}.yml, socratiq.qmd, and HTML integration tests)
- Reconstruct MathJax/KaTeX inline TeX in Study Tools text extraction while stripping mjx-assistive-mml and .chapter-concept-map noise
- Index MLSysBook foldbox callouts and closed <details> summary keywords in create_text_ref.js
- Expand Focused Reader captureParagraph to support list items, figure captions, and blockquotes inside <main>
- Bridge Focused Reader active passage into Study Tools + Context and @ mentions, add Escape close and documentElement theme sync
- Remediate DOM XSS sinks, hardcoded third-party API keys, runtime crash paths, Spaced Repetition bugs, listener leaks, and bundle minification
Add reproducible Epoch AI frontier training compute (@fig-ai-compute-growth) and single-run pretraining energy (@fig-frontier-training-energy) figures to Volume II, Chapter 15, paired with builder-reported anchors at declared measurement boundaries, deterministic CSV extraction and manifest receipts, and structured provenance enforcement across book CSV series and figure cells.
Compact mode (create({layout:'compact', intro})) orders the player as
title, intro, collapsed assumptions, diagram, control bar, narration.
The bar pins with a one-line step title only when it avoids the step's
focus and keep regions; a keyboard-reachable pill links to the narration
otherwise. Speed, Restart and the key legend move into a Speed popover.
Status words, step labels, view switches (Follow/Whole) and intros follow
new lints in check_original_player.py.
Pilots: mixed precision, overlapped fetching, activation checkpointing
(ch08); H100 roofline batching, tiling (ch11). Non-compact players are
unchanged.
dompurify 3.4.17 closes the critical sanitizer advisory. The committed
bundle is rebuilt so the fix ships with it; old and new bundles both load
in headless Chromium with no errors. The low-severity katex advisory
remains in the copies pinned by mermaid and ink-mde.
Each use of ledger now names what the thing is: the scenario and coverage
tables of the data chapter, event timing tables, stopping and latency
budgets, trial and outcome logs, audit logs, test reports, and sets of
fault records. Labels, index keys, printed field names, concept maps, LEGO
class names, and figure text follow, and two chapter covers are relabeled.
Nearby inflated phrasing is plainer.
A new chapter 17, after Release, teaches who answers for the decisions
behind an admitted action, oversight a person can actually perform,
permission in stages, how a detection gap between groups of people
becomes a gap in impact energy, who bears risk without choosing it, data
about people, and the work the machine moves onto people. It opens on the
2023 Cruise collision. The approval record names who approved every
decision and blocks release when a name is missing. Privacy and incident
disclosure move in from Release, and worker burden from Frontier. The
fourth law is now named permission must be earned. The part opener,
roadmap, spine figure, dependency map, appendices, glossary, labs, and
concept maps follow the new order.
Makes room for the responsible physical AI chapter after Release. Pure
rename: the folder, chapter file, its -17- figure and table labels and
fig17_ image files move to 18, and configs, concept-map owners, the
structure table, lab links and figure tooling follow.
Retitle the chapter ML Lifecycle and rename its folder, files, assets,
lab, slides, audit data, and instructor links, with redirects for the
old URLs. Section IDs keep the sec-ml-workflow prefix. Update the ch2
hand-off and the conclusion to name the lifecycle.
Rename the chapter folder, assets, lab, slides, audit data, and staffml
links from 02_ml_systems to 02_system_paradigms, with redirects for the
old URLs. Revise the introduction, notation, conventions, and about
pages, adding the frontmatter stack figures. Add park2024offloading.
New scopes, each with a dated note at the check: headers blank-before
(headings and fenced divs folded into a paragraph lose their anchor) and
unnumbered-target; markup prose-leaks (LaTeX ties, $ closers glued to a
digit or an inline ref, caption-attribute escapes); refs callout-ref-noun
(custom callout refs print a bare number without their noun); notation
registry, vol1 print pin, and forms (now default; vol4 transpose rule);
pdf figure-font-floor (opt-in, printed-point measurement). inline-order gets
its own pre-commit hook so skipping book-check-code no longer hides it. The
notation-consistency bare-B rule ignores bold matrices. vol4 margin
generator: SPRT bounds and the p99 ladder label.
- Execute multi-tier live online verification across all 2,578 unique bibliography entries.
- Fix hallucinated and conflated entries in bibs/references.bib and mlperf-edu/paper/refs.bib.
- Replace broken and redirected URLs with authentic DOIs and canonical HTTPS URLs.
- Enrich bibs/canonical_provenance.json with 491 verified canonical records.
- Reseal bibs/manifest.json with updated cryptographic hashes.
- Add binder/tools/scripts/audit/comprehensive_live_bib_verifier.py with retry logic.
- Verify 100.0% clean pass on all binder check and pytest suites.
Relocate master bibliography (references.bib), cryptographic ledger
(manifest.json), curated external provenance (canonical_provenance.json),
and staging candidate inbox (inbox.bib) into unified bibs/ directory.
Maintain books/references.bib as git-tracked relative symlink to
../bibs/references.bib for seamless Quarto and compiler compatibility.
Update Binder CLI, verification engines, pre-commit guards, codespell
rules, and gateway test fixtures to point to bibs/ paths.
- Run multi-tiered parallel verification engine against Crossref, DataCite, arXiv, OpenLibrary, standards registries, and authoritative web endpoints.
- Seal all 2,454 entries in books/references.bib into binder/tools/bib_verification_manifest.json with cryptographic SHA-256 hashes and authentic provenance.
- Harden binder check bib --scope verify to strictly require status: 'verified', an authentic external source, and a matching SHA-256 hash.
- Prune uncited duplicate and fabricated citations from references.bib and correct cadene2024lerobot repository metadata.
- Add canonical provenance registry and unit tests for hardened verification checks.
- Ch13 deadline counts use Ch4's full response budget (blocking and
interference allowances); the "upper limits" hedge is gone
- Blocking and interference allowances promoted to MLSysIM (AISLE), read by
Ch4 and Ch13, listed in the data sheet (138 rows)
- Ch2 freshness worked example on the running machine's pick-station
conveyor: median fits, the tail misses
- Ch5 transfer sections use the running machine; Cruise war story cut (Ch12
owns it), orphaned figure and uncited bib entries removed
- Ch17 shield section kept as synthesis and tightened; its parent heading now
names what the section argues
SKIP=book-check-bib-verify: the vol4 bib ledger is deferred citation work
(550 pre-existing unverified entries, none added by this commit).
- Ch4 winding time constant from the registry (2.7 ms, was hand-typed 0.5 ms)
- tau_delay defined as the pre-brake delay excluding delta_da in Ch1, Ch2 and
appendix_control; operator MAC is AES-128 CMAC in Ch14 and appendix_systems
- Ch4 handoff grid matches the intent model's 336 ms at 2 Hz
- Ch8 wrist-camera exposure from the registry (5 ms); Ch9 40 ms named as the
chunk-policy inference pass; unsourced percentages reworded
- Unrated checks no longer called certified (Ch11, Ch12)
- Remaining hand-typed values to LEGO; MLSysIM insulation class F limit
- LOs aligned with body (Ch14 LO3); S.P.A glossed at first use
- Margin generator: handshake strip status ready (preface swapped it out)