[Juliusolsson05]

What this is

The full RENDER-layer rewrite per docs/superpowers/specs/2026-07-11-feed-render-layer-rewrite-design.md and its 21-task implementation plan (both in this branch). The DECIDE layer (ownership ledger, view bridge, FeedRenderItem contract) is untouched — every change is in the painter and provider renderer modules.

Headline fixes

• Live code streams highlighted line-by-line with zero remounts. The open-fence path used to rebuild a full Monaco editor + model + LSP document on every streaming token (and remounted again when the fence language arrived a delta late). StreamingCodeBlock seals each completed line into a per-line hljs cache; only the tail line re-tokenizes per delta.
• Streaming markdown stops re-parsing the whole message per delta (SegmentedMarkdown: the prefix through the last closed fence parses once; O(len²) → O(len)).
• Command output is ANSI-aware everywhere (AnsiText on the terminal palette; \r progress rewrites collapse) and exit codes finally display (red exit N badges; Codex silent successes render a $ cmd ✓ record instead of vanishing).
• One card per tool family, identical live and committed (ui/resolve/ + ui/artifacts/): CommandCard, DiffCard (Edit/MultiEdit/apply_patch incl. tinted patch-failure diffs — successes used to render nothing), FileWriteCard (live-highlighted streaming writes), ReadCard, TodoCard, WebCard/ImageGenCard (Codex special tools get committed parity), SlashCommandRow (no more raw <command-name> tag soup), ThinkingBlock (redacted gets an explicit pill), GenericToolCard (the ONE fallback — live MCP tools no longer dump raw partial JSON).
• Feed.tsx 1060 → ~640 lines; the scarred scroll/picker/debug behaviors are ported verbatim into ui/hooks/. Deleted: both dispatch ladders' tool routing, JsonToolRow/ToolUseRow/DiffSlab-consumers, ClaudeRows, Codex exec/patch/stdin rows, both TruncatedOutputRow copies, renderToolUse from the capability contract.
• Claude spawn matching accepts Task alongside Agent (future-proofing; current transcripts emit Agent).

Verification

• tsc clean on both projects; 481/481 tests pass including the bundle + recording corpora (zero divergence — the one rollout change, silent-success results, was absorbed cleanly). The 1 failed test file is the documented pre-existing empty hotkeyBinding suite.
• electron-vite build + remote-client build pass (kit stays browser-pure; Monaco remains desktop-lazy).
• Needs a live eyeball before merge (the plan's phase-gate checklist, spec §9): a streaming multi-fence answer, Edit/apply_patch/Write bursts, a colored failing command, WebSearch/MCP/subagent/slash-command rows, provider switch mid-session, and the scroll behaviors.

🤖 Generated with Claude Code

[Juliusolsson05]

Superseded by #533, which merged the tested combined feed rewrite plus the follow-up streaming and presentation fixes into main.

[Juliusolsson05]

Correction: #533 was reverted by #534 because these changes are unfinished. Reopening this PR for its original review and development lifecycle.

[Juliusolsson05]

Status update: this PR is intentionally back in draft and must not be merged in its current form. The current rendering rewrite has accumulated substantial shape-assumption, reconciliation, and maintainability problems. We are changing the implementation philosophy: first build an unknown/distinctive event-shape collector and a reviewed fixture corpus from real Claude/Codex/OpenCode traffic; then define classified render contracts and modular renderers from that evidence. This branch requires a heavy rebase/reconciliation against current main before any new implementation work, and much of the present painter should be expected to be replaced rather than patched forward.

[Juliusolsson05]

WIP checkpoint pushed from the newer local feed-render-rewrite worktree.

Commit: 1d46769

This preserves the materially newer local implementation on the draft PR so it is reviewable and cannot be lost. It includes the pure presentation projection, projection receipts, stable operation rows, provider extractors, presentation primitives, fixtures, and the later Rules-of-Hooks crash fix.

Validation run before push:

• 12 unit files / 80 tests passed
• 12 renderer files / 46 tests passed
• production build passed
• git diff --check passed

Important: this checkpoint is evidence and salvage material, not an assertion that the current 11.8k-line migration is the architecture we should merge. The PR remains intentionally draft. The next design pass is evidence-first: build the distinctive/unknown shape collector and classified fixture corpus before freezing renderer families or replacing legacy paths.

Provider ownership rule clarified after reviewing this checkpoint:

• Claude and Codex must not share a provider interpretation or dispatch layer.
• Each provider owns decoding its wire shapes, partial-stream behavior, result pairing, and lifecycle into its own explicit adapter.
• Sharing starts only at narrow domain presentation protocols such as code edit, file write, command, read, or search models.
• Shared React code renders those domain protocols; provider adapters map into them.
• A new provider therefore gets an independent decoder and can reuse proven visual protocols without inheriting Claude/Codex assumptions.

The checkpoint should now be reviewed for parts to retain, rewrite, or delete after the shape corpus exists.

[Juliusolsson05]

Superseded by the evidence-first provider-owned renderer rewrite in #555. Phase 9 now records each valuable #524 primitive as ported, selectively ported, reimplemented, or rejected in docs/rendering/legacy-deletion-manifest.md. The global artifact projection and shared provider classifiers are intentionally not carried forward. #555 remains open through Phase 10 and is not being merged yet.