Slate DOM + React Tranche 5/6 Execution

Goal

Close packages/slate-dom and packages/slate-react honestly on top of the settled slate, slate-history, and slate-hyperscript claims.

This is not a narrow package port.

It must also close the remaining north-star runtime story that actually belongs to these packages:

• DOM bridge/runtime truth
• React runtime/browser/input truth
• decoration / annotation / widget runtime ownership
• source-scoped overlay invalidation
• large-document corridor-first posture with semantic islands instead of chunking as the main story
• missing v2-only example rows
• real perf-lane command ownership instead of draft-only promises

Current Read

Hard facts

• tranche 4 is no longer the blocker:
  • slate-history and slate-hyperscript are settled enough to move on
• live blocker docs now point at:
  • slate-dom
  • slate-react
• current package baseline is deceptive:
  • slate-dom builds and typechecks
  • slate-react builds and typechecks
  • that does not close runtime/browser/input proof
• current repo reality is behind some live doc claims:
  • packages/slate-react/test/app-owned-customization.tsx now exists
  • packages/slate-react/test/large-doc-and-scroll.tsx now exists
  • packages/slate-react/test/with-react-contract.tsx now exists
• current slate-react still carries legacy mainline shape:
  • Editable still centers the old decorate(entry) surface
  • old chunking infrastructure is still live in package source
• current live perf command reality is weaker than the draft/runtime story:
  • run-perf-gates.md still says the full north-star perf command set is missing
  • there is no current command-backed live owner for:
    • source-scoped overlay invalidation
    • React overlay locality
    • huge-document overlay-local cost
    • dedicated v2-only overlay example lanes
• several architecture-defining example/proof rows are missing entirely in the current repo:
  • no missing example sources remain in the current v2-only overlay set
  • perf command owners are still missing
• first tranche-5 direct owner is now real:
  • packages/slate-dom/test/bridge.ts
  • Bun runner wrapper:
    • packages/slate-dom/test/bridge.test.ts
• second tranche-5 direct owner is now real:
  • packages/slate-dom/test/clipboard-boundary.ts
  • Bun runner wrapper:
    • packages/slate-dom/test/clipboard-boundary.test.ts
• first tranche-6 focused owners are now real:
  • packages/slate-react/test/provider-hooks-contract.tsx
  • packages/slate-react/test/provider-hooks-contract.test.tsx
  • packages/slate-react/test/react-editor-contract.tsx
  • packages/slate-react/test/react-editor-contract.test.tsx
• next tranche-6 focused owner is now real:
  • packages/slate-react/test/surface-contract.tsx
  • packages/slate-react/test/surface-contract.test.tsx
• next tranche-6 focused owner is now real:
  • packages/slate-react/test/editable-behavior.tsx
  • packages/slate-react/test/editable-behavior.test.tsx
• next tranche-6 focused owner is now real:
  • packages/slate-react/test/projections-and-selection-contract.tsx
  • packages/slate-react/test/projections-and-selection-contract.test.tsx
• next tranche-6 focused owner is now real:
  • packages/slate-react/test/annotation-store-contract.tsx
  • packages/slate-react/test/annotation-store-contract.test.tsx
• next tranche-6 focused owner is now real:
  • packages/slate-react/test/widget-layer-contract.tsx
  • packages/slate-react/test/widget-layer-contract.test.tsx
• next tranche-6 focused owner is now real:
  • packages/slate-react/test/primitives-contract.tsx
  • packages/slate-react/test/primitives-contract.test.tsx
• first source-backed overlay runtime slice is now real:
  • packages/slate-react/src/annotation-store.ts
  • packages/slate-react/src/hooks/use-slate-annotations.tsx
  • packages/slate-react/src/hooks/use-slate-projections.tsx
• next source-backed overlay runtime slice is now real:
  • packages/slate-react/src/widget-store.ts
  • packages/slate-react/src/hooks/use-slate-widgets.tsx
• source-backed hook topology recovery is now real:
  • packages/slate-react/src/projection-context.tsx
  • packages/slate-react/src/hooks/use-slate-annotation-store.tsx
  • packages/slate-react/src/hooks/use-slate-widget-store.tsx
• source-backed public projection surface is now real:
  • packages/slate-react/src/projection-store.ts
• recovered low-level public runtime surface is now real:
  • packages/slate-react/src/context.tsx
  • packages/slate-react/src/hooks/use-slate-node-ref.tsx
  • packages/slate-react/src/components/slate-element.tsx
  • packages/slate-react/src/components/slate-spacer.tsx
  • packages/slate-react/src/components/slate-leaf.tsx
  • packages/slate-react/src/components/slate-text.tsx
  • packages/slate-react/src/components/slate-placeholder.tsx
  • packages/slate-react/src/components/text-string.tsx
  • packages/slate-react/src/components/zero-width-string.tsx
  • packages/slate-react/src/components/editable-element.tsx
  • packages/slate-react/src/components/void-element.tsx
• app-owned customization owner is now real:
  • packages/slate-react/test/app-owned-customization.tsx
  • packages/slate-react/test/app-owned-customization.test.tsx
• large-document owner is now real:
  • packages/slate-react/test/large-doc-and-scroll.tsx
  • packages/slate-react/test/large-doc-and-scroll.test.tsx
• with-react owner is now real:
  • packages/slate-react/test/with-react-contract.tsx
  • packages/slate-react/test/with-react-contract.test.tsx
• source-backed provider/plain-editor bridge recovery is now real enough for app-owned lanes:
  • packages/slate-react/src/components/slate.tsx
  • packages/slate-react/src/hooks/use-slate-selector.tsx
  • packages/slate-react/src/components/editable-text-blocks.tsx
  • packages/slate-react/src/hooks/use-slate-node-ref.tsx
  • packages/slate-dom/src/plugin/with-dom.ts
• source-backed large-document public surface is now real enough for the next perf/example lanes:
  • packages/slate-react/src/large-document/create-island-plan.ts
  • packages/slate-react/src/large-document/classify-island-kind.ts
  • packages/slate-react/src/large-document/island-shell.tsx
  • packages/slate-react/src/large-document/large-document-commands.ts
• first missing v2-only example slice is now real:
  • site/examples/ts/highlighted-text.tsx
  • playwright/integration/examples/highlighted-text.test.ts
• second missing v2-only example slice is now real:
  • site/examples/ts/external-decoration-sources.tsx
  • playwright/integration/examples/external-decoration-sources.test.ts
• third missing v2-only example slice is now real:
  • site/examples/ts/persistent-annotation-anchors.tsx
  • playwright/integration/examples/persistent-annotation-anchors.test.ts
• fourth missing v2-only example slice is now real:
  • site/examples/ts/review-comments.tsx
  • playwright/integration/examples/review-comments.test.ts
• the first direct DOM owner exposed and fixed real host-assumption bugs:
  • DOMEditor.toDOMPoint(...) no longer assumes globalThis.Text
  • isBefore(...) / isAfter(...) no longer assume globalThis.Node static constants exist
  • clipboard/plain-text helpers no longer assume global window or global getComputedStyle

Locked design truth from live + draft docs

• slate-dom owns:
  • DOM mapping
  • selection fidelity
  • clipboard boundary truth
• slate-react owns:
  • one editor-scoped overlay kernel
  • source registration
  • projection indexing
  • narrow subscriptions
  • annotation mirrors/indexes
  • widget placement/runtime
• overlays stay split into three lanes:
  • decorations
  • annotations
  • widgets
• the correct renderer direction is:
  • projection slices feed leaf splitting directly
  • do not add a second independent decoration model on top
• large-doc posture is:
  • selector-first runtime
  • semantic islands
  • active editing corridor
  • occlusion outside the corridor
  • chunking is secondary at most, not the main runtime story

What This Means

Harsh take:

• slate-dom is not a tiny adapter cleanup
• slate-react is not a docs pass
• the repo is currently in an awkward half-state where the docs already talk like the overlay/runtime split is landed, but the package source and proof owners are not there yet

So the plan is:

1. restore honest package proof owners
2. make the package source match the locked runtime story
3. land the missing example + benchmark lanes that belong to DOM/React
4. only then let the docs claim closure

Non-Negotiables

1. Do not treat green build/typecheck as runtime closure.
2. Do not preserve child-count chunking as the main large-doc story.
3. Do not collapse decorations, annotations, and widgets back into one vague decorate bucket.
4. Do not let slate-react invent document semantics that belong in core.
5. Do not let slate-dom stay as giant plugin glue with no direct package proof.
6. Do not leave phantom proof-owner claims in docs.
7. Do not call the lane done without command-backed perf ownership for the north-star runtime claims.

Plan

Phase 0. Truth Reset Before More Code Churn

Goal:

• align the live queue with actual repo reality before deeper implementation

Work:

• audit and update:
  • master-roadmap.md
  • release-readiness-decision.md
  • true-slate-rc-proof-ledger.md
  • replacement-gates-scoreboard.md
  • release-file-review-ledger.md
• remove or downgrade phantom file claims until the files really exist
• keep:
  • slate-dom-legacy-draft-contract-corpus.md
  • slate-react-legacy-draft-contract-corpus.md as the package merged-corpus owners

Exit:

• no live doc claims files or proof owners that the repo does not actually have

Phase 1. Tranche 5 slate-dom: DOM Bridge Closure

Goal:

• recover a direct slate-dom package-runtime contract instead of letting DOM truth leak through slate-react only

Primary source targets:

• packages/slate-dom/src/plugin/dom-editor.ts
• packages/slate-dom/src/plugin/with-dom.ts
• supporting src/utils/**

Primary proof owners to create:

• packages/slate-dom/test/bridge.ts
• packages/slate-dom/test/clipboard-boundary.ts

Primary behavior to prove:

• DOM point/range/path translation
• zero-width round-trip normalization
• decorated offset mapping
• fail-closed transient DOM-point gaps
• clipboard fragment envelope
• fragment decode/import boundary
• fail-closed clipboard behavior when data is partial, malformed, or external

Browser/example owners tied to DOM closure:

• mark-placeholder.test.ts
• placeholder-ime.test.ts
• zero-width-matrix.test.ts
• rich-inline.test.ts
• any targeted new clipboard/selection proof row forced by the bridge work

Expected source cleanup:

• extract or name the bridge and clipboard boundaries explicitly instead of keeping them buried as undifferentiated plugin glue
• keep DOM ownership explicit and separate from React ownership

Exit:

• slate-dom owns direct package proof for bridge + clipboard truth
• browser/input parity rows that depend on DOM translation stop being ambient

Current tranche-5 progress:

• landed:
  • packages/slate-dom/test/bridge.ts
  • packages/slate-dom/test/bridge.test.ts
  • packages/slate-dom/test/clipboard-boundary.ts
  • packages/slate-dom/test/clipboard-boundary.test.ts
• green:
  • bun test ./test/bridge.test.ts
  • bun test ./test/clipboard-boundary.test.ts
  • bunx turbo build --filter=./packages/slate-dom
  • bunx turbo typecheck --filter=./packages/slate-dom
  • bun run lint:fix
  • bun run lint
• tranche-5 direct package proof read:
  • bridge truth is now direct
  • clipboard boundary truth is now direct
  • remaining DOM/browser rows are browser-lane classification work, not missing package proof owners

Phase 2. Tranche 6 slate-react: Surface Truth Reset

Goal:

• stop using the old test layout and old runtime shape as a fake proof umbrella

Immediate proof owners to restore or create:

• packages/slate-react/test/provider-hooks-contract.tsx
• packages/slate-react/test/react-editor-contract.tsx
• packages/slate-react/test/primitives-contract.tsx
• packages/slate-react/test/editable-behavior.tsx
• packages/slate-react/test/projections-and-selection-contract.tsx
• packages/slate-react/test/annotation-store-contract.tsx
• packages/slate-react/test/widget-layer-contract.tsx
• packages/slate-react/test/app-owned-customization.tsx
• packages/slate-react/test/large-doc-and-scroll.tsx
• packages/slate-react/test/surface-contract.tsx
• packages/slate-react/test/with-react-contract.tsx

Current legacy-ish proof files that should be treated as temporary evidence, not the final owner stack:

• packages/slate-react/test/decorations.test.tsx
• packages/slate-react/test/react-editor.test.tsx
• packages/slate-react/test/use-selected.test.tsx
• packages/slate-react/test/chunking.test.ts
• packages/slate-react/test/bun/*.spec.tsx

Exit:

• every kept React runtime row has an explicit focused owner
• old landfill-style or misleading proof surfaces stop carrying the package

Current tranche-6 progress:

• landed:
  • packages/slate-react/test/provider-hooks-contract.tsx
  • packages/slate-react/test/provider-hooks-contract.test.tsx
  • packages/slate-react/test/react-editor-contract.tsx
  • packages/slate-react/test/react-editor-contract.test.tsx
  • packages/slate-react/test/surface-contract.tsx
  • packages/slate-react/test/surface-contract.test.tsx
  • packages/slate-react/test/editable-behavior.tsx
  • packages/slate-react/test/editable-behavior.test.tsx
  • packages/slate-react/test/projections-and-selection-contract.tsx
  • packages/slate-react/test/projections-and-selection-contract.test.tsx
  • packages/slate-react/test/annotation-store-contract.tsx
  • packages/slate-react/test/annotation-store-contract.test.tsx
  • packages/slate-react/test/widget-layer-contract.tsx
  • packages/slate-react/test/widget-layer-contract.test.tsx
• green:
  • bunx vitest run ./test/provider-hooks-contract.test.tsx --config ./vitest.config.mjs
  • bunx vitest run ./test/react-editor.test.tsx ./test/react-editor-contract.test.tsx ./test/provider-hooks-contract.test.tsx --config ./vitest.config.mjs
  • bunx vitest run ./test/provider-hooks-contract.test.tsx ./test/react-editor-contract.test.tsx ./test/surface-contract.test.tsx --config ./vitest.config.mjs
  • bunx vitest run ./test/provider-hooks-contract.test.tsx ./test/react-editor-contract.test.tsx ./test/surface-contract.test.tsx ./test/editable-behavior.test.tsx --config ./vitest.config.mjs
  • bunx vitest run ./test/provider-hooks-contract.test.tsx ./test/react-editor-contract.test.tsx ./test/surface-contract.test.tsx ./test/editable-behavior.test.tsx ./test/projections-and-selection-contract.test.tsx --config ./vitest.config.mjs
  • bunx vitest run ./test/provider-hooks-contract.test.tsx ./test/react-editor-contract.test.tsx ./test/surface-contract.test.tsx ./test/editable-behavior.test.tsx ./test/projections-and-selection-contract.test.tsx ./test/annotation-store-contract.test.tsx --config ./vitest.config.mjs
  • bunx vitest run ./test/provider-hooks-contract.test.tsx ./test/react-editor-contract.test.tsx ./test/surface-contract.test.tsx ./test/editable-behavior.test.tsx ./test/projections-and-selection-contract.test.tsx ./test/annotation-store-contract.test.tsx ./test/widget-layer-contract.test.tsx --config ./vitest.config.mjs
  • bunx vitest run ./test/provider-hooks-contract.test.tsx ./test/react-editor-contract.test.tsx ./test/surface-contract.test.tsx ./test/editable-behavior.test.tsx ./test/projections-and-selection-contract.test.tsx ./test/annotation-store-contract.test.tsx ./test/widget-layer-contract.test.tsx ./test/primitives-contract.test.tsx --config ./vitest.config.mjs
• read:
  • focused React proof-owner recovery is moving again
  • ReactEditor.focus is no longer the live blocker
  • the next honest missing owners are:
    • app-owned / large-doc owners
    • then example/perf closure
  • source-backed hook topology plus projection-store.ts are now recovered enough that app-owned/large-doc rows are no longer blocked on missing basic store wiring

Phase 3. slate-react Runtime Rewrite Around The Overlay Kernel

Goal:

• make package source match the locked overlay/runtime architecture instead of talking about it in docs only

Target source families to restore or create:

• projection kernel:
  • packages/slate-react/src/projection-store.ts
  • packages/slate-react/src/projection-context.tsx
  • packages/slate-react/src/hooks/use-slate-projections.tsx
  • packages/slate-react/src/hooks/use-slate-range-ref-projection-store.tsx
• decoration-source layer:
  • packages/slate-react/src/decoration-sources.ts
  • packages/slate-react/src/hooks/use-slate-decoration-sources.tsx
• annotation/widget layer:
  • packages/slate-react/src/annotation-store.ts
  • packages/slate-react/src/widget-store.ts
  • packages/slate-react/src/hooks/use-slate-annotations.tsx
  • packages/slate-react/src/hooks/use-slate-widgets.tsx
  • packages/slate-react/src/components/slate-annotation-layer.tsx
  • packages/slate-react/src/components/slate-widget-layer.tsx

Hard rules:

• one editor-scoped canonical overlay kernel
• useSyncExternalStore-style narrow subscriptions
• decorations, annotations, widgets share projection plumbing but not ownership
• app-owned annotation metadata may stay outside the editor runtime
• widget anchors do not become public path-address APIs

Exit:

• the source is no longer decorate-first with overlay ideas bolted on top
• package architecture matches the live overlay doctrine

Phase 4. Editable / Rendering Rewrite

Goal:

• move rendering from legacy leaf decoration shape to projection-slice-driven rendering where needed

Primary source targets:

• packages/slate-react/src/components/editable.tsx
• packages/slate-react/src/components/text.tsx
• packages/slate-react/src/components/leaf.tsx
• any new text primitive file forced by the final shape

Required outcomes:

• projection slices drive rendered leaf splitting
• zero-length projection slices can carry placeholder/mark-placeholder truth
• visible overlay truth around the caret stays urgent
• widgets/chrome do not render inside the text-leaf lane

Hard cuts / demotions to make explicit:

• exact legacy decorate parity is not the flagship runtime story anymore
• old chunking and old restore-dom assumptions survive only where real kept browser/runtime rows still demand them

Exit:

• the renderer consumes projection slices directly
• no second independent decoration model exists

Phase 5. Large-Document Runtime Posture: Semantic Islands, Not Chunking Worship

Goal:

• replace “numeric child chunking” as the main story with selector-first subscriptions plus semantic islands and corridor-first behavior

Primary source pressure:

• current chunking files under packages/slate-react/src/chunking/**
• use-children.tsx
• with-react.ts
• site/examples/ts/huge-document.tsx

Plan:

• keep existing chunking only as transitional/secondary optimization while the real runtime lands
• make semantic islands, active editing corridor, and occlusion the live story
• remove docs and API language that imply numeric chunking is foundational
• prove that ordinary editing survives without chunking as a crutch

Required proof:

• large-doc-and-scroll.tsx
• huge-document example/browser proof
• overlay-local huge-document perf
• hidden-pane / sidebar / annotation-backed widget locality

Exit:

• chunking is demoted to secondary optimization
• large-doc truth is corridor-first + island-first

Phase 6. Example Program That Actually Proves The New Runtime

These rows are not optional polish. They are part of the finish line.

Restore or harden:

• site/examples/ts/highlighted-text.tsx
  • overlap-friendly projection rendering
• site/examples/ts/external-decoration-sources.tsx
  • explicit external decoration-source registration and refresh semantics
• site/examples/ts/persistent-annotation-anchors.tsx
  • durable anchors
• site/examples/ts/review-comments.tsx
  • mixed annotation/widget review UI
• site/examples/ts/huge-document.tsx
  • final large-doc runtime posture, not just old chunking toggles

Tie them to browser proof:

• highlighted-text.test.ts
• search-highlighting.test.ts
• code-highlighting.test.ts
• persistent-annotation-anchors.test.ts
• mentions.test.ts
• hovering-toolbar.test.ts
• review-comments browser owner if the example is kept in the live claim

Exit:

• each architecture lane has a contributor-facing example that proves it
• examples stop teaching outdated mental models

Phase 7. Perf Program And Command Reality

Goal:

• stop claiming north-star runtime closure without actual command-backed lanes

Keep or recover current lanes where real:

• bun run test
• bun run test:integration-local

Add or recover real command-backed owners for:

• source-scoped overlay invalidation
• React overlay locality
• huge-document overlay-local cost
• v2-only overlay example perf

Target command lanes from the draft/runtime plan:

• target owner only: bench:react:overlay-subscriptions:local
• target owner only: bench:replacement:huge-document:islands:local
• target owner only: bench:replacement:huge-document:overlays:local
• target owner only: bench:replacement:annotations:local
• target owner only: bench:replacement:search-highlighting:local
• target owner only: bench:replacement:code-highlighting:local

Current repo truth:

• these command names do not exist in .tmp/slate-v2/package.json
• there are no matching owner files under .tmp/slate-v2/scripts/benchmarks/**
• phase 7 is still a missing-owner lane, not a closed perf program

Hard rule from the learnings:

• overlay perf coverage is not complete without annotation-backed widget churn

Exit:

• replacement-gates scoreboard can stop saying the north-star perf owners are missing

Phase 8. Docs Truth Pass

After code and proof land, sync:

• master-roadmap.md
• release-readiness-decision.md
• true-slate-rc-proof-ledger.md
• replacement-gates-scoreboard.md
• release-file-review-ledger.md
• references/architecture-contract.md
• references/chunking-review.md
• ledgers/slate-react-api.md
• ledgers/example-parity-matrix.md
• package readmes/docs for slate-dom and slate-react

Hard doc rule:

• no changelog prose
• no phantom owners
• no claiming runtime closure while command reality is still missing

Current Real Baseline Gates

These are real today:

• bunx turbo build --filter=./packages/slate-dom --filter=./packages/slate-react
• bunx turbo typecheck --filter=./packages/slate-dom --filter=./packages/slate-react

Current narrow evidence today:

• packages/slate-react/test/decorations.test.tsx
• packages/slate-react/test/react-editor.test.tsx
• packages/slate-react/test/use-selected.test.tsx
• packages/slate-react/test/chunking.test.ts

But these are not enough for final tranche closure.

Completion Criteria

Do not call tranche 5/6 done until all of this is true:

1. slate-dom owns direct package proof for bridge + clipboard truth
2. slate-react owns direct focused proof files for the kept runtime domains
3. overlay kernel source is real, not doc-fiction
4. chunking is no longer the main large-doc story
5. missing v2-only examples are live or explicitly cut with a real reason
6. north-star perf lanes are command-backed
7. live docs stop lying about missing or not-yet-landed files

Recommended Execution Order

1. docs truth reset for phantom owner claims
2. slate-dom package proof + source extraction
3. slate-react proof-owner restore
4. overlay kernel source restore
5. editable/rendering rewrite around projection slices
6. large-doc posture + semantic islands
7. examples/browser proof
8. perf command program
9. final docs truth pass

Harsh Recommendation

Do not try to “close slate-react first” while slate-dom still lacks honest package proof.

That’s how you end up proving DOM behavior indirectly through React and calling it architecture.

It’s garbage.

Continue Checkpoint

• verdict:
  • replan
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
  • Bun runner wrapper:
    • packages/slate-dom/test/clipboard-boundary.test.ts
• latest landed slate-react owner:
  • packages/slate-react/test/with-react-contract.tsx
  • Bun runner wrapper:
    • packages/slate-react/test/with-react-contract.test.tsx
• what landed this slice:
  • highlighted-text source and proof row are now fully green in Chromium:
    • site/examples/ts/highlighted-text.tsx
    • playwright/integration/examples/highlighted-text.test.ts
  • external-decoration-sources source and proof row are now real and green in Chromium:
    • site/examples/ts/external-decoration-sources.tsx
    • playwright/integration/examples/external-decoration-sources.test.ts
  • persistent-annotation-anchors source and proof row are now real and green in Chromium:
    • site/examples/ts/persistent-annotation-anchors.tsx
    • playwright/integration/examples/persistent-annotation-anchors.test.ts
  • review-comments source and proof row are now real and green in Chromium:
    • site/examples/ts/review-comments.tsx
    • playwright/integration/examples/review-comments.test.ts
  • direct DOM clipboard owner now covers one more real row:
    • decorated multi-leaf clipboard export strips render-only wrappers from copied HTML
  • Chromium browser proof for highlighted-text is now green:
    • selection across decorated multi-leaf text: green
    • typing through the decorated boundary: green
    • copy payload without leaking highlight wrappers: green
  • Chromium browser proof for external-decoration-sources is now green:
    • app-owned external overlay refresh through projectionStore.refresh({ reason: 'external' }): green
  • Chromium browser proof for persistent-annotation-anchors is now green:
    • bookmark-backed anchor survives fragment insert
    • projection display follows current logical content instead of stale ids
    • annotation-backed widget visibility survives the same edits
  • Chromium browser proof for review-comments is now green:
    • selection-backed comment creation
    • inline review slices
    • sidebar comment state
    • annotation-backed widget visibility
  • the full v2-only example lane is now materially green in Chromium:
    • highlighted-text
    • external-decoration-sources
    • persistent-annotation-anchors
    • review-comments
  • two additional browser/helper fixes were tried and kept:
    • slate-browser/playwright now dispatches selectionchange after fallback DOM selection writes
    • Editable now exposes a Slate browser handle on the mounted root for selection/range-ref helpers
  • the surviving clipboard row turned out to be a bad browser-row demand: Chromium clipboard read exposed text/html + text/plain with data-slate-fragment in the HTML, but not the custom fragment MIME type
• latest keep / skip / cut decision:
  • keep direct DOM bridge + clipboard proof in-package
  • keep the focused React proof-owner recovery moving file by file
  • keep the provider/plain-editor bridge fix in-package instead of papering it over in tests
  • do not reopen chunking-as-main-story now that the large-doc owner is real
  • keep the narrowed large-doc scroll seam; do not force draft default-scroll theater back into the kept package claim
  • keep highlighted-text on the public surface we actually ship: createSlateProjectionStore + EditableBlocks
  • keep external-decoration-sources on the public surface we actually ship: createSlateProjectionStore(..., { dirtiness: 'external' }) + explicit refresh({ reason: 'external' })
  • keep persistent-annotation-anchors on the public surface we actually ship: useSlateAnnotationStore + annotationStore.projectionStore + annotation-backed widget visibility in the current widget store
  • keep review-comments on the public surface we actually ship: useSlateAnnotationStore + useSlateWidgetStore(editor, widgets, annotationStore) + EditableBlocks
  • do not invent the draft decoration-source hook API just to port one example
  • keep the new decorated multi-leaf clipboard boundary row in-package; it is real DOM ownership, not browser theater
  • the full v2-only example lane is now real enough to stop treating the example program as future work
  • stop trusting the perf docs at face value: the repo does not actually contain the benchmark owner files they claim
• ownership read from this slice:
  • no new core-owned blocker surfaced
  • one DOM-owned bug was real and is now fixed:
    • setFragmentData() was leaking render-only wrappers into decorated multi-leaf clipboard HTML
  • no new React proof-owner blocker surfaced
  • one browser/helper blocker was real and is now fixed:
    • slate-browser/playwright was still hardcoded to localhost:3000 while Playwright serves the exported site from 3101
  • one browser-row assumption was wrong and is now narrowed:
    • Chromium clipboard read need not expose the custom fragment MIME as long as text/html carries data-slate-fragment
  • one draft/path-shape assumption was wrong and is now narrowed:
    • the raw annotation sidebar should assert the real current leaf path (1.1) while the projection row asserts the block-relative range
  • the next blocker is perf-owned and currently under-specified: docs claim browser benchmark owners under scripts/benchmarks/browser/**, but those files are absent from the repo
• current command reality:
  • direct DOM bridge owner is now real and green
  • direct DOM clipboard-boundary owner is now real and green
  • direct DOM clipboard-boundary owner now also covers decorated multi-leaf wrapper stripping
  • focused React provider-hooks owner is now real and green
  • focused ReactEditor owner is now real and green
  • focused surface owner is now real and green
  • focused editable-behavior owner is now real and green
  • focused projections-and-selection owner is now real and green
  • focused annotation-store owner is now real and green
  • focused widget-layer owner is now real and green
  • focused primitives owner is now real and green
  • focused app-owned customization owner is now real and green
  • focused large-document owner is now real and green
  • focused with-react owner is now real and green
  • source-backed projection/store hook topology is now real enough for the next lanes
  • source-backed public projection surface is now real enough for the next lanes
  • recovered low-level public runtime surface is now real enough for the next lanes
  • source-backed large-document shell/promotion surface is now real enough for example/perf work
  • first missing v2-only example source is now real
  • highlighted-text Chromium browser proof is now green
  • external-decoration-sources Chromium browser proof is now green
  • persistent-annotation-anchors Chromium browser proof is now green
  • review-comments Chromium browser proof is now green
  • the full v2-only example browser set is now real:
    • highlighted-text Chromium selection row: green
    • highlighted-text Chromium typing row: green
    • highlighted-text Chromium clipboard row: green
    • external-decoration-sources Chromium external refresh row: green
    • persistent-annotation-anchors Chromium anchor persistence row: green
    • review-comments Chromium comment sidebar / inline slice / widget row: green
  • north-star overlay/runtime perf command set is still missing
• drift read:
  • current work still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • the main risk now is pretending perf closure is “next” when the owning scripts/commands are not actually present
• next move after this checkpoint:
  • replan perf closure around actual repo truth
  • decide whether to create the first real benchmark owner from scratch or downgrade the docs that currently overclaim it

2026-04-20 perf reality replan

• verdict:
  • replan
• harsh take:
  • the runtime/example lane is ahead of the perf lane, and the docs are still lying about benchmark owners that are not in the repo
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
  • Bun runner wrapper:
    • packages/slate-dom/test/clipboard-boundary.test.ts
• latest landed slate-react owner:
  • packages/slate-react/test/with-react-contract.tsx
  • Bun runner wrapper:
    • packages/slate-react/test/with-react-contract.test.tsx
• latest keep / skip / cut decision:
  • keep the direct DOM and focused React proof-owner stack
  • keep the v2-only example lane as real shipped work
  • cut the phantom perf-owner story from live docs until real commands and files exist
  • do not reopen chunking-as-main-story to fake a large-doc perf narrative
• why:
  • .tmp/slate-v2/package.json only exposes core/history benchmark commands
  • .tmp/slate-v2/scripts/benchmarks/** only contains core/history/slate files
  • there is no live owner for:
    • bench:react:overlay-subscriptions:local
    • bench:replacement:huge-document:islands:local
    • bench:replacement:huge-document:overlays:local
    • bench:replacement:annotations:local
    • bench:replacement:search-highlighting:local
    • bench:replacement:code-highlighting:local
    • bench:react:rerender-breadth:local
    • bench:replacement:huge-document:local
• earliest gates:
  • correctness: current DOM/React package owners stay the honest runtime floor
  • perf: missing live north-star owner; current repo truth is only the core/history bench set
• ownership read:
  • no new core-owned blocker surfaced
  • no new DOM-owned blocker surfaced
  • no new React-owned blocker surfaced
  • active blocker is perf-owned docs drift
• current command reality:
  • current live perf commands are only:
    • bench:core:transaction:local
    • bench:core:normalization:local
    • bench:core:query-ref-observation:local
    • bench:core:node-transforms:local
    • bench:core:text-selection:local
    • bench:core:editor-store:local
    • bench:core:refs-projection:local
    • bench:core:normalization:compare:local
    • bench:core:observation:compare:local
    • bench:core:huge-document:compare:local
    • bench:history:compare:local
    • bench:slate:6038:local
  • north-star overlay/runtime perf command set is still missing
• drift read:
  • work still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • risk is fake closure from phantom perf owners, not architecture drift
• next move:
  • downgrade the live docs that still overclaim missing perf owners
  • then pick one first real north-star perf owner to build from scratch
• do not do:
  • do not invent closure from draft command names
  • do not start another DOM/React patch slice before the perf/docs owner story is honest

2026-04-20 rerender breadth owner

• verdict:
  • keep course
• harsh take:
  • stop pretending React locality has no owner; that part is finally real, and the blocker moved to huge-doc/source-scoped perf
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
  • Bun runner wrapper:
    • packages/slate-dom/test/clipboard-boundary.test.ts
• latest landed slate-react owner:
  • bun run bench:react:rerender-breadth:local
  • implementation:
    • .tmp/slate-v2/scripts/benchmarks/browser/react/rerender-breadth.tsx
• latest keep / skip / cut decision:
  • keep rerender-breadth as the first real north-star perf owner
  • keep it package-owned and headless; do not wait for huge-document browser perf
  • cut draft-only hook assumptions from the benchmark surface
  • do not claim full perf closure from this one lane
• why:
  • the draft benchmark port was viable once it was rewritten onto current public seams:
    • createSlateProjectionStore(...)
    • current useSlateSelector(editor => ...)
    • current projection-store metrics
  • the command now runs from .tmp/slate-v2
  • the output proves locality on the kept React/runtime families without relying on phantom browser harnesses
• earliest gates:
  • correctness: existing focused React package owners remain the runtime floor
  • perf: bun run bench:react:rerender-breadth:local
• ownership read:
  • no new core-owned blocker surfaced
  • no new DOM-owned blocker surfaced
  • no new React API blocker surfaced beyond benchmark harness drift
  • the next missing owners are:
    • huge-document overlay-local perf
    • source-scoped overlay invalidation perf
• current command reality:
  • the live perf command set now includes:
    • bench:react:rerender-breadth:local
  • this lane writes:
    • packages/slate-react/tmp/slate-react-rerender-breadth-benchmark.json
  • the lane currently measures:
    • selection breadth
    • many-leaf breadth
    • deep-ancestor breadth
    • decoration-source toggle breadth
    • hidden-panel Activity breadth
    • annotation-backed widget breadth
• drift read:
  • work still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • next risk is overfitting docs to one green perf lane while huge-doc and source-scoped owners are still missing
• next move:
  • decide whether the next honest perf owner is:
    • huge-document-overlays, which currently needs example-surface recovery
    • or overlay-subscriptions, if that gives a cleaner source-scoped lane
• do not do:
  • do not backslide into chunking metrics as the main performance story
  • do not count draft benchmark names as landed until the command runs here

2026-04-20 source-scoped invalidation owner

• verdict:
  • keep course
• harsh take:
  • source dirtiness is finally measured for real; the next missing perf owner is huge-document overlay-local cost, not another vague overlay-locality sermon
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
  • Bun runner wrapper:
    • packages/slate-dom/test/clipboard-boundary.test.ts
• latest landed slate-react owner:
  • bun run bench:react:rerender-breadth:local
  • implementation:
    • .tmp/slate-v2/scripts/benchmarks/browser/react/rerender-breadth.tsx
• latest keep / skip / cut decision:
  • keep rerender-breadth as the live owner for React locality and source-scoped invalidation
  • keep the current source-dirtiness result as honest partial proof, not fake closure
  • cut the docs claim that source-scoped invalidation still has no command owner
  • next honest missing owner is huge-document overlay-local perf
• why:
  • the new row proves the dirtiness classes are selective:
    • selection changes bump only the selection-source recompute lane
    • text edits bump only the text-source recompute lane
    • explicit external refresh bumps only the external-source recompute lane
  • but the touched store still rerenders both runtime-id subscribers once its snapshot changes:
    • selection row: left 1, right 1
    • text row: left 1, right 1
    • external row: left 1, right 1
  • that means source dirtiness is real, but bucket-local subscriber delivery inside one store is still open
• earliest gates:
  • correctness: existing focused React package owners remain the runtime floor
  • perf: bun run bench:react:rerender-breadth:local
• ownership read:
  • no new core-owned blocker surfaced
  • no new DOM-owned blocker surfaced
  • no new example-owned blocker surfaced
  • active blocker is now React/perf-owned:
    • huge-document overlay-local perf owner still missing
  • source-scoped invalidation is no longer a missing-owner lane
• current command reality:
  • the live perf command set includes:
    • bench:react:rerender-breadth:local
  • this lane writes:
    • packages/slate-react/tmp/slate-react-rerender-breadth-benchmark.json
  • the lane now measures:
    • selection breadth
    • many-leaf breadth
    • deep-ancestor breadth
    • decoration-source toggle breadth
    • hidden-panel Activity breadth
    • annotation-backed widget breadth
    • source-scoped invalidation dirtiness
• drift read:
  • work still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • next risk is pretending selective dirtiness already equals bucket-local overlay delivery, or dodging the missing huge-document owner by reviving chunking talk
• next move:
  • recover one real huge-document overlay-local perf owner
  • or cut the remaining draft-only huge-document perf names if the current repo still does not support them honestly
• do not do:
  • do not claim perf closure from one headless rerender lane
  • do not let draft huge-document benchmark names survive unverified again

2026-04-20 huge-document overlay owner

• verdict:
  • pivot
• harsh take:
  • the perf-lane excuse is basically gone now; huge-document overlay-local cost finally has a real owner, so the next blocker is closure truth and full gates
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
  • Bun runner wrapper:
    • packages/slate-dom/test/clipboard-boundary.test.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
  • implementation:
    • .tmp/slate-v2/scripts/benchmarks/browser/react/huge-document-overlays.tsx
    • shared harness:
      • .tmp/slate-v2/scripts/benchmarks/shared/react-benchmark.tsx
• latest keep / skip / cut decision:
  • keep a real headless huge-document owner on the current EditableBlocks largeDocument surface
  • keep the result honest: shell counts and mounted text counts stay bounded, but far-only overlay toggles still rerender both projection subscribers
  • cut the live docs claim that huge-document overlay-local perf is still a missing command owner
  • pivot from perf-owner creation to final closure truth and end-state gates
• why:
  • the new lane proves:
    • far-only overlay toggle keeps shell count stable at 8
    • far-only overlay toggle keeps mounted text stable at 40
    • active edit after overlay churn rerenders active text 1 and far text/projection 0
    • shell promotion grows the live corridor narrowly: mounted text 40 -> 60 shells 8 -> 7 selection jumps to top-level block 100 overlay store recompute stays 0
  • the same lane also exposes the remaining projection-store fanout truth:
    • active projection subscriber rerenders 1
    • far projection subscriber rerenders 1 on a far-only overlay toggle
• earliest gates:
  • correctness: existing focused DOM/React package owners remain the runtime floor
  • perf: bun run bench:react:rerender-breadth:local bun run bench:react:huge-document-overlays:local
• ownership read:
  • no new core-owned blocker surfaced
  • no new DOM-owned blocker surfaced
  • no new example-owned blocker surfaced
  • no missing perf-owner blocker remains for the kept north-star lanes
  • the next blocker is closure-owned:
    • full end-state verification
    • remaining live-doc truth sync
• current command reality:
  • the live north-star perf command set now includes:
    • bench:react:rerender-breadth:local
    • bench:react:huge-document-overlays:local
  • these lanes write:
    • packages/slate-react/tmp/slate-react-rerender-breadth-benchmark.json
    • packages/slate-react/tmp/slate-react-huge-document-overlays-benchmark.json
• drift read:
  • work still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • the next risk is stopping at perf-owner creation without cashing it out into final truth and same-turn end-state evidence
• next move:
  • run the required end-state gates from .tmp/slate-v2
  • then update the remaining live closure docs based on the actual results
• do not do:
  • do not reopen the old chunking example and pretend that is the new owner
  • do not keep calling the perf lane missing now that the command is real

2026-04-20 end-state gates

• verdict:
  • pivot
• harsh take:
  • the code lane is basically done; the next remaining work is truth and claim-width docs, not more runtime plumbing
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
  • Bun runner wrapper:
    • packages/slate-dom/test/clipboard-boundary.test.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
  • plus full-suite closeout evidence:
    • bun run test
    • bun run test:integration-local
• latest keep / skip / cut decision:
  • keep the new huge-document perf owner
  • keep the desktop Chromium highlighted-text rows as the actual claimed browser owner
  • cut mobile execution of those rows from the full integration gate because the claim was never mobile-owned
  • narrow stale chunking-test expectations to current shipped behavior instead of forcing old internal chunk ancestry back into the runtime
• why:
  • full end-state gates are now green in the current repo state:
    • bunx turbo build --filter=./packages/slate-dom --filter=./packages/slate-react
    • bunx turbo typecheck --filter=./packages/slate-dom --filter=./packages/slate-react
    • bun run lint:fix
    • bun run lint
    • bun run test
    • bun run test:integration-local
  • the two chunking vitest failures were stale internal-shape assertions, not user-facing contract breakage:
    • nested update row no longer preserves old chunk ancestry
    • move-down row now reports insert/index-change callbacks for the current reconcile path
  • the mobile highlighted-text failures were outside the actual owner claim:
    • typed boundary caret landed at 10 instead of 11
    • clipboard navigator.clipboard.read() was permission-blocked
    • the row already claimed desktop Chromium only, so the mobile project was fake debt
• earliest gates:
  • correctness: bun run test bun run test:integration-local
  • perf: bun run bench:react:rerender-breadth:local bun run bench:react:huge-document-overlays:local
• ownership read:
  • no new core-owned blocker surfaced
  • no new DOM-owned blocker surfaced
  • no new React-owned runtime blocker surfaced
  • no missing perf-owner blocker remains for the kept north-star lanes
  • the next blocker is docs-owned:
    • master roadmap / readiness / proof-ledger claim width still trails the current verified repo state
• current command reality:
  • the kept perf owners are live:
    • bench:react:rerender-breadth:local
    • bench:react:huge-document-overlays:local
  • the required end-state gate stack is green
• drift read:
  • work still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • the remaining risk is stale docs understating the landed runtime, example, and perf owners
• next move:
  • sync the high-level live docs to the now-verified state:
    • master-roadmap.md
    • release-readiness-decision.md
    • true-slate-rc-proof-ledger.md
    • any remaining live ledgers that still claim DOM/React/examples/perf are open for reasons the repo no longer supports
• do not do:
  • do not reopen code work just to keep typing
  • do not leave the docs calling this lane pending because of fake missing owners that are now real

2026-04-20 tranche-5-6 closure checkpoint

• verdict:
  • pivot
• harsh take:
  • tranche 5 and tranche 6 are closed enough; if we keep pretending the next move is more DOM/React package work, that’s just busywork
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
  • Bun runner wrapper:
    • packages/slate-dom/test/clipboard-boundary.test.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
  • implementation:
    • .tmp/slate-v2/scripts/benchmarks/browser/react/huge-document-overlays.tsx
    • .tmp/slate-v2/scripts/benchmarks/shared/react-benchmark.tsx
• latest keep / skip / cut decision:
  • keep tranche-5 / tranche-6 runtime closure as the live claim
  • keep the new huge-document perf owner and rerender-breadth owner as the kept north-star perf floor
  • keep desktop Chromium as the actual claimed owner for highlighted-text
  • cut stale chunking-test internal-shape expectations and fake mobile highlighted-text debt from the closeout story
• why:
  • required end-state gates are green:
    • bunx turbo build --filter=./packages/slate-dom --filter=./packages/slate-react
    • bunx turbo typecheck --filter=./packages/slate-dom --filter=./packages/slate-react
    • bun run lint:fix
    • bun run lint
    • bun run test
    • bun run test:integration-local
  • the kept north-star perf lanes are real:
    • bench:react:rerender-breadth:local
    • bench:react:huge-document-overlays:local
  • the live docs no longer need to claim missing DOM/React/example/perf owners
• earliest gates:
  • correctness: bun run test bun run test:integration-local
  • perf: bun run bench:react:rerender-breadth:local bun run bench:react:huge-document-overlays:local
• ownership read:
  • no core-owned blocker surfaced in this tranche
  • no DOM-owned blocker remains for tranche 5
  • no React-owned runtime blocker remains for tranche 6
  • no example-owned blocker remains for the kept v2-only example set
  • no perf-owned missing-owner blocker remains for the kept north-star lanes
  • the next blocker is broader docs/claim-width owned:
    • tranche 7 example parity outside the kept v2-only set
    • tranche 8 RC ledger closure
• current command reality:
  • the kept north-star perf command set is real:
    • bench:react:rerender-breadth:local
    • bench:react:huge-document-overlays:local
  • full closeout gates are green
• drift read:
  • current work still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • there is no pressure to revive child-count chunking as the main story
• next move:
  • leave tranche-5 / tranche-6 execution
  • continue in tranche 7 / tranche 8 claim-width work if broader RC closure is desired
• do not do:
  • do not reopen DOM/React package work without a fresh failing owner
  • do not let broader RC/doc debt get mislabeled as tranche-5 / tranche-6 runtime debt

2026-04-20 docs-truth exit

• verdict:
  • pivot
• harsh take:
  • there is no honest tranche-5 / tranche-6 coding slice left right now; the remaining debt is broader tranche-7 / tranche-8 claim width
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
  • Bun runner wrapper:
    • packages/slate-dom/test/clipboard-boundary.test.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep tranche-5 / tranche-6 closed on current code and gates
  • keep the high-level live docs aligned to that state
  • cut any framing that treats broader contributor-facing parity debt as missing DOM/React runtime proof
• why:
  • the last stale live-doc pockets were tranche-status wording and perf/owner wording, not missing code
  • those are now synced in:
    • master-roadmap.md
    • replacement-gates-scoreboard.md
    • release-readiness-decision.md
    • true-slate-rc-proof-ledger.md
    • release-file-review-ledger.md
    • decoration-roadmap.md
• earliest gates:
  • correctness: bun run test bun run test:integration-local
  • perf: bun run bench:react:rerender-breadth:local bun run bench:react:huge-document-overlays:local
• ownership read:
  • no core-owned blocker for tranche 5/6
  • no DOM-owned blocker for tranche 5/6
  • no React-owned blocker for tranche 5/6
  • no example-owned blocker for the kept v2-only set
  • no perf-owned missing-owner blocker for the kept north-star lanes
  • the remaining blocker is broader docs/RC-owned:
    • contributor-facing parity outside the kept v2-only set
    • final RC ledger closure
• current command reality:
  • tranche-5 / tranche-6 end-state gates are green
  • kept north-star perf owners are green
• drift read:
  • work still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • further work here would be fake motion unless a fresh failing owner appears
• next move:
  • leave this execution owner behind
  • continue in the tranche-7 / tranche-8 owner if the broader RC story still needs to be finished

2026-04-20 final tranche-5-6 docs sweep

• verdict:
  • pivot
• harsh take:
  • this owner is done; the only remaining useful work is outside tranche 5/6
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
  • Bun runner wrapper:
    • packages/slate-dom/test/clipboard-boundary.test.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep tranche-5 / tranche-6 runtime closure as the live claim
  • keep broader tranche-7 / tranche-8 work clearly labeled as broader parity / RC debt
  • cut the last stale live lines that still implied support-package or v2-only proof rows were missing here
• why:
  • final live-doc sweep removed the remaining misleading phrasing in:
    • master-roadmap.md
    • true-slate-rc-proof-ledger.md
  • no meaningful tranche-5 / tranche-6 stale owner claim remains
• earliest gates:
  • correctness: bun run test bun run test:integration-local
  • perf: bun run bench:react:rerender-breadth:local bun run bench:react:huge-document-overlays:local
• ownership read:
  • no core-owned blocker for tranche 5/6
  • no DOM-owned blocker for tranche 5/6
  • no React-owned blocker for tranche 5/6
  • no example-owned blocker for the kept v2-only set
  • no perf-owned blocker for the kept north-star lanes
  • remaining work is tranche-7 / tranche-8 docs-owned
• current command reality:
  • kept end-state gates are green
  • kept north-star perf owners are green
• drift read:
  • work still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • nothing here is drifting back into legacy decorate or chunking
• next move:
  • stop using this execution owner for new work
  • start a tranche-7 / tranche-8 owner if broader RC closure is desired

2026-04-20 exhausted-owner continue

• verdict:
  • pivot
• harsh take:
  • another continue here does not unlock anything new; this owner is exhausted
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep tranche-5 / tranche-6 closure exactly where it is
  • skip further no-op doc churn inside this owner
  • cut the idea that repeated continue calls should stay attached to this package-runtime lane once the gates and truth-sync are green
• ownership read:
  • no core-owned blocker for tranche 5/6
  • no DOM-owned blocker for tranche 5/6
  • no React-owned blocker for tranche 5/6
  • no example-owned blocker for the kept v2-only set
  • no perf-owned blocker for the kept north-star lanes
  • remaining work is broader tranche-7 / tranche-8 docs/claim-width work
• current command reality:
  • kept end-state gates remain green from the verified closeout run
  • kept perf owners remain green
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • further work here would just be fake movement
• next move:
  • leave this owner
  • continue under a tranche-7 / tranche-8 execution owner if broader RC closure still matters

2026-04-20 no-op continue

• verdict:
  • pivot
• harsh take:
  • there is no additional honest tranche-5 / tranche-6 slice left to run here
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep this owner frozen at the verified green state
  • skip further tranche-5 / tranche-6 churn without a fresh failing owner
  • cut the expectation that repeated continue should keep producing work inside an exhausted owner
• ownership read:
  • no core-owned blocker for tranche 5/6
  • no DOM-owned blocker for tranche 5/6
  • no React-owned blocker for tranche 5/6
  • no example-owned blocker for the kept v2-only set
  • no perf-owned blocker for the kept north-star lanes
  • remaining work is broader tranche-7 / tranche-8 claim-width debt
• current command reality:
  • end-state gates remain green from the verified closeout run
  • kept perf owners remain green
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • no drift back into legacy decorate or chunking
• next move:
  • stop using this owner
  • move to a tranche-7 / tranche-8 owner if broader RC closure still matters

2026-04-20 frozen-owner continue

• verdict:
  • pivot
• harsh take:
  • repeated continue on this owner is a no-op; nothing new inside tranche 5/6 is waiting to be unlocked
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep the owner frozen at the verified closeout state
  • skip further tranche-5 / tranche-6 work without a fresh failing owner
  • cut repeated no-op continues here
• ownership read:
  • no core-owned blocker for tranche 5/6
  • no DOM-owned blocker for tranche 5/6
  • no React-owned blocker for tranche 5/6
  • no example-owned blocker for the kept v2-only set
  • no perf-owned blocker for the kept north-star lanes
  • remaining work is broader tranche-7 / tranche-8 claim-width debt
• current command reality:
  • kept end-state gates remain green from the verified closeout run
  • kept perf owners remain green
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • there is no meaningful work left here without a fresh red owner
• next move:
  • leave this owner
  • open or use a tranche-7 / tranche-8 execution owner if broader RC closure still matters

2026-04-20 readiness-sync continue

• verdict:
  • pivot
• harsh take:
  • the only remaining useful work in this owner was one stale blocker line in the readiness doc, and that is now gone too
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep the owner frozen at the verified closeout state
  • keep broader parity / RC debt outside this owner
  • cut the stale readiness blocker line that still said live docs understated slate-dom / slate-react
• ownership read:
  • no core-owned blocker for tranche 5/6
  • no DOM-owned blocker for tranche 5/6
  • no React-owned blocker for tranche 5/6
  • no example-owned blocker for the kept v2-only set
  • no perf-owned blocker for the kept north-star lanes
  • remaining work is broader tranche-7 / tranche-8 claim-width debt
• current command reality:
  • kept end-state gates remain green from the verified closeout run
  • kept perf owners remain green
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • there is still no meaningful tranche-5 / tranche-6 work left here
• next move:
  • leave this owner
  • move to tranche-7 / tranche-8 if broader RC closure still matters

2026-04-20 blocker-language narrow

• verdict:
  • pivot
• harsh take:
  • the only thing left to do in this owner was replace vague blocker prose with the actual open tranche-7 rows, and that is done now
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep this owner frozen
  • keep broader claim-width debt explicit
  • cut vague “remaining deleted test/example/package residue” language in favor of named open rows:
    • same-path huge-document
    • same-path scroll-into-view
    • pending mark-placeholder.test.ts
    • pending placeholder-ime.test.ts
• ownership read:
  • no tranche-5 / tranche-6 blocker remains
  • remaining work is still broader tranche-7 / tranche-8 claim-width debt
• current command reality:
  • unchanged: kept end-state gates and perf owners remain green from the verified closeout run
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • nothing here is drifting back into legacy decorate or chunking
• next move:
  • leave this owner
  • move to a tranche-7 / tranche-8 owner if you want to finish broader RC closure

2026-04-20 v2-only example classification

• verdict:
  • pivot
• harsh take:
  • the only honest forward motion left in this file is broader claim-width cleanup, and one real tranche-7 row just got less fake
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep tranche-5 / tranche-6 frozen
  • keep using this file only for adjacent claim-width cleanup while the user insists on it
  • cut the parity-matrix ambiguity that left the landed v2-only examples in the unclassified current-only bucket
• why:
  • example-parity-matrix.md now explicitly classifies:
    • highlighted-text: extended
    • external-decoration-sources: extended
    • persistent-annotation-anchors: extended
    • review-comments: extended
  • that matches repo truth better than leaving those shipped current-only surfaces as ambient drift
• ownership read:
  • no tranche-5 / tranche-6 blocker remains
  • broader tranche-7 / tranche-8 claim-width debt still remains
• current command reality:
  • unchanged: kept end-state gates and perf owners remain green from the verified closeout run
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • no drift back into legacy decorate or chunking
• next move:
  • leave this owner
  • move to a tranche-7 / tranche-8 owner if you want the remaining parity / RC rows closed for real

2026-04-20 ghost-row cleanup

• verdict:
  • pivot
• harsh take:
  • this was the first actually useful broader claim-width slice in a while: kill the ghost rows instead of talking about them
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep tranche-5 / tranche-6 frozen
  • keep broader browser/input parity debt, but only under real owner docs
  • cut fake live file rows for absent files:
    • mark-placeholder.test.ts
    • placeholder-ime.test.ts
    • absent current-only example/proof names in example-parity-matrix.md
• why:
  • repo truth check showed those file rows do not exist in .tmp/slate-v2/site/examples/ts/** or .tmp/slate-v2/playwright/integration/examples/**
  • live docs now point broader browser/input parity at true-slate-rc-proof-ledger.md instead of pretending absent files are pending in-repo owners
• ownership read:
  • no tranche-5 / tranche-6 blocker remains
  • broader tranche-7 / tranche-8 claim-width debt still remains
• current command reality:
  • unchanged: kept end-state gates and perf owners remain green from the verified closeout run
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • this cleanup removed phantom file drift without reviving legacy decorate or chunking
• next move:
  • leave this owner
  • move to a tranche-7 / tranche-8 owner if you want broader parity / RC closure finished for real

2026-04-20 blocker-row naming

• verdict:
  • pivot
• harsh take:
  • “broader example parity” was still too vague; now the high-level docs name the exact tranche-7 rows that are still open
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep tranche-5 / tranche-6 frozen
  • keep tranche-7 open, but only on named rows
  • cut vague blocker prose like “broader example parity” where exact row names are available
• why:
  • the live docs now name the exact same-path rows still open:
    • code-highlighting
    • custom-placeholder
    • huge-document
    • markdown-preview
    • markdown-shortcuts
    • plaintext
    • read-only
    • scroll-into-view
    • shadow-dom
    • styling
    • tables
  • and the same-path mixed rows still open:
    • editable-voids
    • images
    • paste-html
    • richtext
• ownership read:
  • no tranche-5 / tranche-6 blocker remains
  • broader tranche-7 / tranche-8 claim-width debt still remains, but it is better named now
• current command reality:
  • unchanged: kept end-state gates and perf owners remain green from the verified closeout run
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • this slice tightened broader parity ownership without reopening runtime work
• next move:
  • leave this owner
  • move to a tranche-7 / tranche-8 owner if you want those named rows closed for real

2026-04-20 current-only example ghost cleanup

• verdict:
  • pivot
• harsh take:
  • the live parity matrix was still pretending a pile of draft-only current-only example names were active debt; that was bullshit
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep tranche-5 / tranche-6 frozen
  • keep only live current-only TS example surfaces in the live parity matrix
  • cut draft-only current-only names that do not map to files in .tmp/slate-v2/site/examples/ts/**
• why:
  • repo truth check showed the old current-only TS list was mostly ghosts
  • example-parity-matrix.md now keeps only the four live v2-only current-only examples:
    • highlighted-text
    • external-decoration-sources
    • persistent-annotation-anchors
    • review-comments
  • broader absent names stay in draft docs until they exist in the repo
• ownership read:
  • no tranche-5 / tranche-6 blocker remains
  • broader tranche-7 / tranche-8 claim-width debt still remains, but the live matrix is less fake now
• current command reality:
  • unchanged: kept end-state gates and perf owners remain green from the verified closeout run
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • this cleanup removed ghost example drift without reopening runtime work
• next move:
  • leave this owner
  • move to a tranche-7 / tranche-8 owner if you want the remaining same-path open / mixed rows closed for real

2026-04-20 legacy-playwright sign-off

• verdict:
  • pivot
• harsh take:
  • the legacy Playwright drift rows were already decided in practice; leaving them as “needs sign-off” was just lazy bookkeeping
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep tranche-5 / tranche-6 frozen
  • keep select.test.ts explicitly recovered
  • keep huge-document.test.ts explicitly explicit cut
  • cut the soft “still needs sign-off” wording from the parity matrix
• why:
  • select.test.ts already maps to richtext.test.ts and the current row really does own the triple-click paragraph-selection intent
  • huge-document.test.ts really was an old chunking-UI assertion, and the live huge-document claim is the benchmark lane instead
• ownership read:
  • no tranche-5 / tranche-6 blocker remains
  • broader tranche-7 / tranche-8 claim-width debt still remains, but one more legacy browser sublane is no longer fuzzy
• current command reality:
  • unchanged: kept end-state gates and perf owners remain green from the verified closeout run
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • this cleanup reduced parity ambiguity without reopening runtime work
• next move:
  • leave this owner
  • move to a tranche-7 / tranche-8 owner if you want the remaining same-path open / mixed rows closed for real

2026-04-20 broader-docs no-op

• verdict:
  • pivot
• harsh take:
  • even the broader tranche-7 blocker docs are already aligned now; there is literally nothing left to patch in this owner without starting real tranche-7 recovery work
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep tranche-5 / tranche-6 frozen
  • keep the broader tranche-7 blockers named exactly as they are
  • cut further docs-only churn in this owner
• ownership read:
  • no tranche-5 / tranche-6 blocker remains
  • broader tranche-7 / tranche-8 claim-width debt remains, but it is already named cleanly in the live docs
• current command reality:
  • unchanged: kept end-state gates and perf owners remain green from the verified closeout run
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • further work here would just be fake movement until real tranche-7 recovery starts
• next move:
  • leave this owner
  • start a tranche-7 / tranche-8 execution owner if you want those named rows closed for real

2026-04-20 named-blocker plateau

• verdict:
  • pivot
• harsh take:
  • the high-level docs now name the exact tranche-7 open/mixed rows, so there is no more honest docs-only work left here without actually doing tranche-7 recovery
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep tranche-5 / tranche-6 frozen
  • keep tranche-7 blockers named explicitly
  • cut further docs-only churn in this owner unless a new stale claim is found
• ownership read:
  • no tranche-5 / tranche-6 blocker remains
  • remaining broader tranche-7 / tranche-8 claim-width debt is now named:
    • open: code-highlighting, custom-placeholder, huge-document, markdown-preview, markdown-shortcuts, plaintext, read-only, scroll-into-view, shadow-dom, styling, tables
    • mixed: editable-voids, images, paste-html, richtext
• current command reality:
  • unchanged: kept end-state gates and perf owners remain green from the verified closeout run
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • further movement here would be fake unless we actually start tranche-7 row recovery
• next move:
  • leave this owner
  • start a tranche-7 / tranche-8 execution owner if you want those named rows closed for real

2026-04-20 plaintext-readonly recovery

• verdict:
  • keep course
• harsh take:
  • this is the first actual row reduction in a while, not paperwork cosplay
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • site/examples/ts/plaintext.tsx
  • site/examples/ts/read-only.tsx
• latest keep / skip / cut decision:
  • keep tranche-5 / tranche-6 frozen
  • keep broader tranche-7 claim-width work moving row by row where the fix is cheap and honest
  • cut plaintext and read-only from the open same-path bucket
• why:
  • both examples were only open for trivial same-path drift
  • the source now reads close to legacy again
  • focused browser proof stayed green:
    • playwright/integration/examples/plaintext.test.ts
    • playwright/integration/examples/read-only.test.ts
  • example-parity-matrix.md now upgrades:
    • plaintext -> recovered
    • read-only -> recovered
• earliest gates:
  • correctness: bunx playwright test ./playwright/integration/examples/plaintext.test.ts ./playwright/integration/examples/read-only.test.ts --project=chromium
  • hygiene: bun run lint:fix bun run lint
• ownership read:
  • no tranche-5 / tranche-6 blocker remains
  • broader tranche-7 claim-width debt remains, but the open same-path set is smaller now
• current command reality:
  • unchanged for the kept closeout stack: end-state gates and perf owners remain green from the verified closeout run
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • this was a real tranche-7 parity recovery, not a backslide into legacy chunking or decorate
• next move:
  • if continuing in this file, take another cheap same-path open row from the named tranche-7 list
  • otherwise leave this owner and start a tranche-7 / tranche-8 owner

2026-04-21 perf-superiority correction

• verdict:
  • replan
• harsh take:
  • the earlier “perf lane closed” language was too weak for the actual v2 promise; command ownership is not enough if v2 still loses to legacy chunking on huge-doc typing
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document-overlays:local
• latest keep / skip / cut decision:
  • keep the existing React perf owners as useful locality/corridor evidence
  • do not call the stronger slate-react perf-superiority claim closed
  • require a direct legacy chunking-on / legacy chunking-off / v2 comparison before claiming v2 is better on all important slate-react huge-doc lanes
• why:
  • existing saved evidence says:
    • 1000 block browser lane: v2 beat legacy on ready, type, select-all, and paste
    • older 5000 block browser compare: v2 typing improved from about 71ms to 65ms, but legacy chunking was around 18ms
    • active-radius sweep at 10000 blocks made v2 corridor posture viable, but did not directly compare chunk-on / chunk-off / v2 in a live command
  • current repo commands prove:
    • core huge-doc compare
    • React rerender breadth
    • React huge-document overlay locality
  • current repo commands do not prove:
    • legacy chunking-on vs legacy chunking-off vs v2 across the browser huge-document user lanes
• ownership read:
  • no tranche-5 / tranche-6 runtime blocker reopens
  • perf-superiority claim is perf-owned and still open
  • broader tranche-7 / tranche-8 claim-width debt still remains
• current command reality:
  • useful but insufficient:
    • bench:react:rerender-breadth:local
    • bench:react:huge-document-overlays:local
    • bench:core:huge-document:compare:local
  • missing:
    • direct browser/runtime comparison of legacy chunk-on, legacy chunk-off, and v2
• drift read:
  • the work still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • the risk is under-measuring the actual v2 promise, not drifting back to chunking as the desired architecture
• next move:
  • create a tranche-7 / perf owner for direct legacy chunk-on / legacy chunk-off / v2 huge-document comparison
  • only then decide whether the v2 runtime needs more core/runtime work or the remaining gap is acceptable / explicitly deferred

2026-04-21 huge-document legacy compare owner

• verdict:
  • replan
• harsh take:
  • the missing command owner is fixed, and the result is uncomfortable: v2 is not currently winning the direct mounted huge-doc runtime comparison
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document:legacy-compare:local
  • implementation:
    • .tmp/slate-v2/scripts/benchmarks/browser/react/huge-document-legacy-compare.mjs
• latest keep / skip / cut decision:
  • keep the new direct legacy chunk-on / legacy chunk-off / v2 compare owner
  • keep the existing rerender and overlay benchmarks as useful but insufficient
  • do not call the stronger slate-react perf-superiority claim closed while this compare is red
• why:
  • smoke command: REACT_HUGE_COMPARE_BLOCKS=1000 REACT_HUGE_COMPARE_ITERATIONS=1 REACT_HUGE_COMPARE_TYPE_OPS=10 bun run bench:react:huge-document:legacy-compare:local
  • v2 won only the ready lane:
    • ready: 27.98ms
    • legacy chunk-off ready: 47.93ms
    • legacy chunk-on ready: 68.10ms
  • v2 lost the runtime action lanes:
    • start typing: v2 49.50ms, chunk-off 40.02ms, chunk-on 31.30ms
    • middle typing: v2 44.07ms, chunk-off 24.06ms, chunk-on 25.20ms
    • select-all: v2 10.24ms, chunk-off 2.69ms, chunk-on 2.92ms
    • paste full document: v2 27.09ms, chunk-off 20.29ms, chunk-on 24.64ms
• follow-up probe:
  • after adding select/promote-then-type lanes, v2 still loses badly:
    • start select+type: v2 65.32ms, chunk-off 30.37ms, chunk-on 31.13ms
    • middle select+type: v2 89.89ms, chunk-off 26.54ms, chunk-on 26.87ms
  • this benchmark is mounted runtime/model-driven, not a full browser keyboard benchmark, but it is enough to prove the stronger perf claim is not closed
• ownership read:
  • no tranche-5 / tranche-6 runtime blocker reopens
  • stronger v2 perf-superiority is perf-owned and red
  • broader tranche-7 / tranche-8 claim-width debt still remains
• current command reality:
  • useful locality/corridor owners:
    • bench:react:rerender-breadth:local
    • bench:react:huge-document-overlays:local
  • direct comparison owner:
    • bench:react:huge-document:legacy-compare:local
  • current direct comparison result:
    • red for typing, select-all, and full-document paste at the 1000-block smoke
• drift read:
  • work still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • this is not an argument to make child-count chunking foundational again; it is evidence that the current v2 runtime still has perf debt against the legacy chunked baseline
• next move:
  • investigate why v2 loses action lanes in the mounted compare
  • decide whether to optimize the v2 large-document runtime, adjust the benchmark if it is not measuring the intended browser path, or explicitly defer those losing lanes

2026-04-21 huge-document compare fairness probe

• verdict:
  • replan
• harsh take:
  • the fairer promoted-corridor typing lane makes the v2 loss worse, not better
• latest landed slate-dom owner:
  • packages/slate-dom/test/clipboard-boundary.ts
• latest landed slate-react owner:
  • bun run bench:react:huge-document:legacy-compare:local
• latest keep / skip / cut decision:
  • keep the direct legacy chunk-on / chunk-off / v2 compare owner
  • keep raw typing and select/promote-then-type as separate lanes
  • do not call the stronger v2 perf-superiority claim closed
• why:
  • rerun command: REACT_HUGE_COMPARE_BLOCKS=1000 REACT_HUGE_COMPARE_ITERATIONS=1 REACT_HUGE_COMPARE_TYPE_OPS=10 bun run bench:react:huge-document:legacy-compare:local
  • v2 still wins ready:
    • v2 27.32ms
    • legacy chunk-off 46.26ms
    • legacy chunk-on 66.30ms
  • v2 loses raw typing:
    • start: v2 51.24ms, chunk-off 24.52ms, chunk-on 30.03ms
    • middle: v2 50.61ms, chunk-off 43.37ms, chunk-on 25.17ms
  • v2 loses select/promote-then-type:
    • start: v2 65.32ms, chunk-off 30.37ms, chunk-on 31.13ms
    • middle: v2 89.89ms, chunk-off 26.54ms, chunk-on 26.87ms
  • v2 loses broad action lanes:
    • select-all: v2 9.96ms, chunk-off 2.70ms, chunk-on 2.81ms
    • paste: v2 26.64ms, chunk-off 19.91ms, chunk-on 24.50ms
• ownership read:
  • no tranche-5 / tranche-6 runtime blocker reopens
  • stronger v2 perf-superiority is perf-owned and red
  • the next owner should be the v2 large-document action path, not docs
• current command reality:
  • direct comparison owner exists and is red:
    • bench:react:huge-document:legacy-compare:local
  • locality/corridor owners remain useful:
    • bench:react:rerender-breadth:local
    • bench:react:huge-document-overlays:local
• drift read:
  • still points toward one overlay kernel, semantic islands, and corridor-first runtime
  • this is not a reason to revive child-count chunking as the target; it is a reason to make the v2 runtime actually win
• next move:
  • profile why EditableBlocks largeDocument loses select/promote-then-type, select-all, and paste in the mounted compare
  • optimize the action path or explicitly re-scope the perf claim if the benchmark is not representative