[pseudoseed]

Merges 1153 upstream commits (pingdotgg/t3code up to 67eebaca23, "fix(server): a stopped run no longer shrinks thread history to one row (pingdotgg#31266)") into the fork and bumps PseudoCode to 0.0.78. Every fork feature survived the merge.

Fork changes kept

• All 130 fork-added files are present at HEAD. The 4 files the fork deleted are still deleted (T3Wordmark x2, T3Mark.svg, AnalyticsService.test.ts).
• Analytics stay off (PR 57). Upstream's new ProviderUsageAnalytics writes through the fork's no-op AnalyticsService, so it sends nothing.
• User-visible text upstream added in the fork's files says PseudoCode / Cloud Connect (cli/up.ts, cloudLinkState.ts, replay fixtures).
• Upstream's Effect 4.1 upgrade was carried into fork code: SecureRandom.uuidV4, NodeCrypto.layer in tests, Effect.fn where lint asks, deterministic service keys.

Fixed during the sync

• Extra Codex accounts sharing a home. Upstream began reading CODEX_HOME from the host environment. With it set, every extra Codex account resolved to that one home and their logins merged. A host CODEX_HOME now applies to the default instance only. New test "applies a host CODEX_HOME to the default instance only" fails without the fix.
• run-suites.sh crashed on a non-ASCII vendored path; it now lists files with core.quotePath=false and skips .repos/.

DROPPED lines from fork-delta-check.py, accepted

27 lines, all explained by an upstream change that does the same job:

• DesktopServerExposure: the fork's cached state became upstream's per-read readCurrentState; the fork's behaviour is folded in.
• RemoteOpenTargets: the fork's cache is replaced by upstream's equivalent; the fork's TTL export is kept.
• ActivitySummaries: thread type narrowed to Pick<OrchestrationV2ThreadShell, "id" | "latestRunId"> for upstream's new shell type.
• UsageLimitSources: apiFor moved into a cliproxyHub helper.
• ThreadLineageWorkflowRow disabled: upstream removed it (fix(clients): stop loading thread families when opening chats pingdotgg/t3code#26762); unloaded members are now enabled.
• The rest are test lines rewritten for upstream's new shapes.

Behaviour changes from upstream

• iOS Live Activity (APNs) now shortens the summary to fit Apple's payload limit before dropping rows; the fork used to drop rows only. The fork test asserts all 5 rows survive.
• Forgejo PR search follows upstream's paging and keeps going when a page has no matches.

Tests

• vp run check:changed, vp run lint (whole repo), typecheck for all 40 packages, and knip pass.
• run-suites.sh: every package passes except:
  • GuestPage browser test (distance 2074) and build-npm-launcher fail on this Mac on clean upstream/main too.
  • apps/web runtime.test.ts times out only under full-suite load; it passes alone (3/3).
• CodexDriver.test.ts 16/16, Kiro replay 17/17.

Live check

Against real data (73 threads copied from a trimmed dev database).

• Web (Browser panel): About shows 0.0.78 and "PseudoCode sends no usage data, traces, or crash reports off your machine." Usage shows priced cost for 1,626 Claude sessions. Pull requests lists Forgejo PRs from git.pseudoseed.com. A Replay (dev) showcase turn streamed, stopped for approval, and finished after Approve. Every page request went to localhost only.
• iOS (iPhone 18 Pro simulator, dev client built for this checkout): paired with the dev server and listed the real threads. Usage → Cost shows $21,949.84; PRs lists GitHub and Forgejo PRs; Settings → About → Diagnostics shows only local startup crashes and has no crash-report upload setting.

Merge with a merge commit; the release tags point at these commits.

🤖 Generated with Claude Code

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ The exact PR base did not have a successful artifact. Baseline uses the latest successful main measurement shown below.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	5.0 KiB	5.0 KiB	−29 B (−0.6%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.2 KiB	−29 B (−2.3%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	21.6 KiB	21.5 KiB	−41 B (−0.2%)	24.4 KiB	✅
Codex	Live turn messages	2	1	−1 (−50.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.9 KiB	21.9 KiB	0 B (0.0%)	24.4 KiB	✅
Claude	Live turn messages	1	1	0 (0.0%)	8	✅

Baseline: 5e1d484 · PR result: 4c57eec · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.