docs(upstream): record the 2026-08-27 T3 review batch by rynfar · Pull Request #112 · pylon-code/pylon

rynfar · 2026-08-27T05:22:55Z

Records the upstream review covering 860caaa60..33b650a5b (4 commits, 4 change sets) and advances reviewed-through to 33b650a5b3b27382b35d2182dec6b22438c3da56.

All four were adopted: #108 (Clerk UI pin removal), #109 (unsettledAt thread ordering), #110 (release CI), #111 (public preview DMG downloads).

The batch note records two things a later session would otherwise have to rediscover:

Why U2's migration is 047. Pylon retired id 36 and renumbered 037–040, so upstream's 043 collided with 043_ProjectionThreadSessionLifecycle.
Why U3's concurrency policy flipped. The review first recommended keeping Pylon's cancel-in-progress on schedule, citing 8dd0d6e06. That rationale had already expired — 124630c3f moved every workflow onto GitHub-hosted runners the day before, so the Blacksmith queue that motivated it can no longer form. The entry records the correction so the next reader does not re-derive it from a stale comment.

The deferred register was empty going in, had nothing to re-evaluate in Phase 2.5, and remains empty — this batch opened no new entries.

Model: Claude Opus 5. Harness: Claude Code.

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

github-actions · 2026-08-27T05:28:24Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.5 KiB	+234 B (+1.7%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−4 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+238 B (+3.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	57.0 KiB	+1.5 KiB (+2.7%)	66.4 KiB	✅
Codex	Live turn messages	9	11	+2 (+22.2%)	21	✅
Claude	Total thread wire	13.3 KiB	13.5 KiB	+178 B (+1.3%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−5 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+183 B (+2.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	57.9 KiB	+1.5 KiB (+2.6%)	66.4 KiB	✅
Claude	Live turn messages	10	11	+1 (+10.0%)	21	✅

Baseline: 3ba1857 · PR result: e6e7067 · 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: 109.4 KiB
Claude decoded thread snapshot: 110.0 KiB

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

rynfar · 2026-08-27T05:22:55Z

Records the upstream review covering 860caaa60..33b650a5b (4 commits, 4 change sets) and advances reviewed-through to 33b650a5b3b27382b35d2182dec6b22438c3da56.

All four were adopted: #108 (Clerk UI pin removal), #109 (unsettledAt thread ordering), #110 (release CI), #111 (public preview DMG downloads).

The batch note records two things a later session would otherwise have to rediscover:

Why U2's migration is 047. Pylon retired id 36 and renumbered 037–040, so upstream's 043 collided with 043_ProjectionThreadSessionLifecycle.
Why U3's concurrency policy flipped. The review first recommended keeping Pylon's cancel-in-progress on schedule, citing 8dd0d6e06. That rationale had already expired — 124630c3f moved every workflow onto GitHub-hosted runners the day before, so the Blacksmith queue that motivated it can no longer form. The entry records the correction so the next reader does not re-derive it from a stale comment.

The deferred register was empty going in, had nothing to re-evaluate in Phase 2.5, and remains empty — this batch opened no new entries.

Model: Claude Opus 5. Harness: Claude Code.

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

github-actions · 2026-08-27T05:28:24Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.5 KiB	+234 B (+1.7%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−4 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+238 B (+3.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	57.0 KiB	+1.5 KiB (+2.7%)	66.4 KiB	✅
Codex	Live turn messages	9	11	+2 (+22.2%)	21	✅
Claude	Total thread wire	13.3 KiB	13.5 KiB	+178 B (+1.3%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−5 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+183 B (+2.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	57.9 KiB	+1.5 KiB (+2.6%)	66.4 KiB	✅
Claude	Live turn messages	10	11	+1 (+10.0%)	21	✅

Baseline: 3ba1857 · PR result: e6e7067 · 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: 109.4 KiB
Claude decoded thread snapshot: 110.0 KiB

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

rynfar · 2026-08-27T05:22:55Z

Records the upstream review covering 860caaa60..33b650a5b (4 commits, 4 change sets) and advances reviewed-through to 33b650a5b3b27382b35d2182dec6b22438c3da56.

All four were adopted: #108 (Clerk UI pin removal), #109 (unsettledAt thread ordering), #110 (release CI), #111 (public preview DMG downloads).

The batch note records two things a later session would otherwise have to rediscover:

Why U2's migration is 047. Pylon retired id 36 and renumbered 037–040, so upstream's 043 collided with 043_ProjectionThreadSessionLifecycle.
Why U3's concurrency policy flipped. The review first recommended keeping Pylon's cancel-in-progress on schedule, citing 8dd0d6e06. That rationale had already expired — 124630c3f moved every workflow onto GitHub-hosted runners the day before, so the Blacksmith queue that motivated it can no longer form. The entry records the correction so the next reader does not re-derive it from a stale comment.

The deferred register was empty going in, had nothing to re-evaluate in Phase 2.5, and remains empty — this batch opened no new entries.

Model: Claude Opus 5. Harness: Claude Code.

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

github-actions · 2026-08-27T05:28:24Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.5 KiB	+234 B (+1.7%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−4 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+238 B (+3.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	57.0 KiB	+1.5 KiB (+2.7%)	66.4 KiB	✅
Codex	Live turn messages	9	11	+2 (+22.2%)	21	✅
Claude	Total thread wire	13.3 KiB	13.5 KiB	+178 B (+1.3%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−5 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+183 B (+2.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	57.9 KiB	+1.5 KiB (+2.6%)	66.4 KiB	✅
Claude	Live turn messages	10	11	+1 (+10.0%)	21	✅

Baseline: 3ba1857 · PR result: e6e7067 · 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: 109.4 KiB
Claude decoded thread snapshot: 110.0 KiB

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

rynfar · 2026-08-27T05:22:55Z

Records the upstream review covering 860caaa60..33b650a5b (4 commits, 4 change sets) and advances reviewed-through to 33b650a5b3b27382b35d2182dec6b22438c3da56.

All four were adopted: #108 (Clerk UI pin removal), #109 (unsettledAt thread ordering), #110 (release CI), #111 (public preview DMG downloads).

The batch note records two things a later session would otherwise have to rediscover:

Why U2's migration is 047. Pylon retired id 36 and renumbered 037–040, so upstream's 043 collided with 043_ProjectionThreadSessionLifecycle.
Why U3's concurrency policy flipped. The review first recommended keeping Pylon's cancel-in-progress on schedule, citing 8dd0d6e06. That rationale had already expired — 124630c3f moved every workflow onto GitHub-hosted runners the day before, so the Blacksmith queue that motivated it can no longer form. The entry records the correction so the next reader does not re-derive it from a stale comment.

The deferred register was empty going in, had nothing to re-evaluate in Phase 2.5, and remains empty — this batch opened no new entries.

Model: Claude Opus 5. Harness: Claude Code.

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

github-actions · 2026-08-27T05:28:24Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.5 KiB	+234 B (+1.7%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−4 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+238 B (+3.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	57.0 KiB	+1.5 KiB (+2.7%)	66.4 KiB	✅
Codex	Live turn messages	9	11	+2 (+22.2%)	21	✅
Claude	Total thread wire	13.3 KiB	13.5 KiB	+178 B (+1.3%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−5 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+183 B (+2.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	57.9 KiB	+1.5 KiB (+2.6%)	66.4 KiB	✅
Claude	Live turn messages	10	11	+1 (+10.0%)	21	✅

Baseline: 3ba1857 · PR result: e6e7067 · 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: 109.4 KiB
Claude decoded thread snapshot: 110.0 KiB

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

rynfar · 2026-08-27T05:22:55Z

Records the upstream review covering 860caaa60..33b650a5b (4 commits, 4 change sets) and advances reviewed-through to 33b650a5b3b27382b35d2182dec6b22438c3da56.

All four were adopted: #108 (Clerk UI pin removal), #109 (unsettledAt thread ordering), #110 (release CI), #111 (public preview DMG downloads).

The batch note records two things a later session would otherwise have to rediscover:

Why U2's migration is 047. Pylon retired id 36 and renumbered 037–040, so upstream's 043 collided with 043_ProjectionThreadSessionLifecycle.
Why U3's concurrency policy flipped. The review first recommended keeping Pylon's cancel-in-progress on schedule, citing 8dd0d6e06. That rationale had already expired — 124630c3f moved every workflow onto GitHub-hosted runners the day before, so the Blacksmith queue that motivated it can no longer form. The entry records the correction so the next reader does not re-derive it from a stale comment.

The deferred register was empty going in, had nothing to re-evaluate in Phase 2.5, and remains empty — this batch opened no new entries.

Model: Claude Opus 5. Harness: Claude Code.

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

github-actions · 2026-08-27T05:28:24Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.5 KiB	+234 B (+1.7%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−4 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+238 B (+3.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	57.0 KiB	+1.5 KiB (+2.7%)	66.4 KiB	✅
Codex	Live turn messages	9	11	+2 (+22.2%)	21	✅
Claude	Total thread wire	13.3 KiB	13.5 KiB	+178 B (+1.3%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−5 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+183 B (+2.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	57.9 KiB	+1.5 KiB (+2.6%)	66.4 KiB	✅
Claude	Live turn messages	10	11	+1 (+10.0%)	21	✅

Baseline: 3ba1857 · PR result: e6e7067 · 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: 109.4 KiB
Claude decoded thread snapshot: 110.0 KiB

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

rynfar · 2026-08-27T05:22:55Z

Records the upstream review covering 860caaa60..33b650a5b (4 commits, 4 change sets) and advances reviewed-through to 33b650a5b3b27382b35d2182dec6b22438c3da56.

All four were adopted: #108 (Clerk UI pin removal), #109 (unsettledAt thread ordering), #110 (release CI), #111 (public preview DMG downloads).

The batch note records two things a later session would otherwise have to rediscover:

Why U2's migration is 047. Pylon retired id 36 and renumbered 037–040, so upstream's 043 collided with 043_ProjectionThreadSessionLifecycle.
Why U3's concurrency policy flipped. The review first recommended keeping Pylon's cancel-in-progress on schedule, citing 8dd0d6e06. That rationale had already expired — 124630c3f moved every workflow onto GitHub-hosted runners the day before, so the Blacksmith queue that motivated it can no longer form. The entry records the correction so the next reader does not re-derive it from a stale comment.

The deferred register was empty going in, had nothing to re-evaluate in Phase 2.5, and remains empty — this batch opened no new entries.

Model: Claude Opus 5. Harness: Claude Code.

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

github-actions · 2026-08-27T05:28:24Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.5 KiB	+234 B (+1.7%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−4 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+238 B (+3.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	57.0 KiB	+1.5 KiB (+2.7%)	66.4 KiB	✅
Codex	Live turn messages	9	11	+2 (+22.2%)	21	✅
Claude	Total thread wire	13.3 KiB	13.5 KiB	+178 B (+1.3%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−5 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+183 B (+2.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	57.9 KiB	+1.5 KiB (+2.6%)	66.4 KiB	✅
Claude	Live turn messages	10	11	+1 (+10.0%)	21	✅

Baseline: 3ba1857 · PR result: e6e7067 · 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: 109.4 KiB
Claude decoded thread snapshot: 110.0 KiB

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

rynfar · 2026-08-27T05:22:55Z

Records the upstream review covering 860caaa60..33b650a5b (4 commits, 4 change sets) and advances reviewed-through to 33b650a5b3b27382b35d2182dec6b22438c3da56.

All four were adopted: #108 (Clerk UI pin removal), #109 (unsettledAt thread ordering), #110 (release CI), #111 (public preview DMG downloads).

The batch note records two things a later session would otherwise have to rediscover:

Why U2's migration is 047. Pylon retired id 36 and renumbered 037–040, so upstream's 043 collided with 043_ProjectionThreadSessionLifecycle.
Why U3's concurrency policy flipped. The review first recommended keeping Pylon's cancel-in-progress on schedule, citing 8dd0d6e06. That rationale had already expired — 124630c3f moved every workflow onto GitHub-hosted runners the day before, so the Blacksmith queue that motivated it can no longer form. The entry records the correction so the next reader does not re-derive it from a stale comment.

The deferred register was empty going in, had nothing to re-evaluate in Phase 2.5, and remains empty — this batch opened no new entries.

Model: Claude Opus 5. Harness: Claude Code.

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

github-actions · 2026-08-27T05:28:24Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.5 KiB	+234 B (+1.7%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−4 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+238 B (+3.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	57.0 KiB	+1.5 KiB (+2.7%)	66.4 KiB	✅
Codex	Live turn messages	9	11	+2 (+22.2%)	21	✅
Claude	Total thread wire	13.3 KiB	13.5 KiB	+178 B (+1.3%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−5 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+183 B (+2.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	57.9 KiB	+1.5 KiB (+2.6%)	66.4 KiB	✅
Claude	Live turn messages	10	11	+1 (+10.0%)	21	✅

Baseline: 3ba1857 · PR result: e6e7067 · 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: 109.4 KiB
Claude decoded thread snapshot: 110.0 KiB

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

rynfar · 2026-08-27T05:22:55Z

Records the upstream review covering 860caaa60..33b650a5b (4 commits, 4 change sets) and advances reviewed-through to 33b650a5b3b27382b35d2182dec6b22438c3da56.

All four were adopted: #108 (Clerk UI pin removal), #109 (unsettledAt thread ordering), #110 (release CI), #111 (public preview DMG downloads).

The batch note records two things a later session would otherwise have to rediscover:

Why U2's migration is 047. Pylon retired id 36 and renumbered 037–040, so upstream's 043 collided with 043_ProjectionThreadSessionLifecycle.
Why U3's concurrency policy flipped. The review first recommended keeping Pylon's cancel-in-progress on schedule, citing 8dd0d6e06. That rationale had already expired — 124630c3f moved every workflow onto GitHub-hosted runners the day before, so the Blacksmith queue that motivated it can no longer form. The entry records the correction so the next reader does not re-derive it from a stale comment.

The deferred register was empty going in, had nothing to re-evaluate in Phase 2.5, and remains empty — this batch opened no new entries.

Model: Claude Opus 5. Harness: Claude Code.

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

github-actions · 2026-08-27T05:28:24Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.5 KiB	+234 B (+1.7%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−4 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+238 B (+3.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	57.0 KiB	+1.5 KiB (+2.7%)	66.4 KiB	✅
Codex	Live turn messages	9	11	+2 (+22.2%)	21	✅
Claude	Total thread wire	13.3 KiB	13.5 KiB	+178 B (+1.3%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−5 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.6 KiB	+183 B (+2.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.4 KiB	57.9 KiB	+1.5 KiB (+2.6%)	66.4 KiB	✅
Claude	Live turn messages	10	11	+1 (+10.0%)	21	✅

Baseline: 3ba1857 · PR result: e6e7067 · 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: 109.4 KiB
Claude decoded thread snapshot: 110.0 KiB

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

docs(upstream): record the 2026-08-27 T3 review batch
9d1f3cf

github-actionsBot added vouch:trusted PR author is trusted by repo permissions or the VOUCHED list. size:M labels Aug 27, 2026

docs(upstream): record the pre-merge adversarial review
e6e7067

github-actionsBot added size:L and removed size:M labels Aug 27, 2026

rynfar merged commit 59a9c46 into pylonAug 27, 2026
15 checks passed

rynfar deleted the docs/upstream-review-2026-08-27 branch August 27, 2026 06:17