docs(agents): record the 2026-08-30 mobile batch and advance the cursor by rynfar · Pull Request #234 · pylon-code/pylon

rynfar · 2026-09-01T16:18:55Z

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and writes the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. The ledger had been sitting two days behind the integration.

The finding worth carrying forward

Upstream landed #8793 (12:22), #8237 (13:35) and #8614 (18:08) within six hours on 2026-08-30, so every later commit carries #8793's composer restructure in its tree as context. It got re-litigated in every single cherry-pick.

It is two independent changes, and treating them as one decision was the mistake:

ComposerSurface — animated borderRadius, absolute glass layer, bounded collapsed radius. Touches nothing in the toolbar. Adopted, so Pylon's file is now structurally identical to upstream and those hunks should stop conflicting.
The toolbar row — declined permanently. ComposerToolbarScroller is upstream's own component and they still ship it; they stopped using it because their toolbar holds four controls. Pylon's holds sixteen, because ControlPillMenu does not exist upstream at all.

Register maintenance

DEF-9 retired — adopted as fix(web): unify activity logs and composer banners #221, once review established that #8734, not #8693, is the correct base (#8733 restores the pre-#8693 tree exactly, so the revert and re-land cancel).
DEF-7 re-checked, still waiting on both premises. Nothing touched desktop-macos-preview.yml, and its 2026-11-01 floor is unreached.
DEF-8 re-checked, and its blocking condition has cleared: Expo SDK 57 merged as eff44d0cf. Only its 2026-09-05 floor remains, four days out.
DEF-10 opened, and it is deliberately unusual. C-19 (#8843, the composer attachment menu) was approved in the same direction as the rest of the batch, but never started. It sits before the new cursor, so no future commit range would ever surface it again. The register is the only thing keeping it visible.

On advancing the cursor with C-19 unfinished

The ledger's own rule is that reviewed-through tracks the newest commit for which every candidate has a decision, not an integration. C-19 has a decision. Advancing is correct by that definition — but it is exactly the case where a cursor move can bury real work, which is what DEF-10 exists to prevent.

Reviewed and integrated with Claude Opus 5 in Claude Code.

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

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and adds the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. Records the finding worth carrying forward: upstream landed #8793, #8237 and #8614 within six hours on 2026-08-30, so each later commit carries #8793's composer restructure as context and it was re-litigated in every cherry-pick. It is two independent changes. ComposerSurface was adopted, so Pylon now matches upstream there and those hunks should stop conflicting. The toolbar row is declined permanently, because ComposerToolbarScroller is upstream's own component that they stopped using at four controls and Pylon has sixteen. Register maintenance: DEF-9 retired, adopted as #221. DEF-7 re-checked and still waiting on both premises. DEF-8 re-checked and its blocking condition has cleared - Expo SDK 57 merged as eff44d0 - leaving only its 2026-09-05 floor. Opens DEF-10 for C-19, which is unusual and deliberate: it is approved work that was never started, and it sits before the new cursor, so no future commit range would surface it. The register is the only thing that keeps it visible.

github-actions · 2026-09-01T16:25:35Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.7 KiB	13.6 KiB	−179 B (−1.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.8 KiB	6.6 KiB	−179 B (−2.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.7 KiB	57.2 KiB	−1.5 KiB (−2.5%)	66.4 KiB	✅
Codex	Live turn messages	11	10	−1 (−9.1%)	21	✅
Claude	Total thread wire	13.5 KiB	13.6 KiB	+53 B (+0.4%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−2 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+55 B (+0.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.0 KiB	58.1 KiB	+44 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	9	10	+1 (+11.1%)	21	✅

Baseline: c908bb2 · PR result: ef7340e · 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: 109.5 KiB
Claude decoded thread snapshot: 110.2 KiB

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

rynfar · 2026-09-01T16:18:55Z

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and writes the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. The ledger had been sitting two days behind the integration.

The finding worth carrying forward

Upstream landed #8793 (12:22), #8237 (13:35) and #8614 (18:08) within six hours on 2026-08-30, so every later commit carries #8793's composer restructure in its tree as context. It got re-litigated in every single cherry-pick.

It is two independent changes, and treating them as one decision was the mistake:

ComposerSurface — animated borderRadius, absolute glass layer, bounded collapsed radius. Touches nothing in the toolbar. Adopted, so Pylon's file is now structurally identical to upstream and those hunks should stop conflicting.
The toolbar row — declined permanently. ComposerToolbarScroller is upstream's own component and they still ship it; they stopped using it because their toolbar holds four controls. Pylon's holds sixteen, because ControlPillMenu does not exist upstream at all.

Register maintenance

DEF-9 retired — adopted as fix(web): unify activity logs and composer banners #221, once review established that #8734, not #8693, is the correct base (#8733 restores the pre-#8693 tree exactly, so the revert and re-land cancel).
DEF-7 re-checked, still waiting on both premises. Nothing touched desktop-macos-preview.yml, and its 2026-11-01 floor is unreached.
DEF-8 re-checked, and its blocking condition has cleared: Expo SDK 57 merged as eff44d0cf. Only its 2026-09-05 floor remains, four days out.
DEF-10 opened, and it is deliberately unusual. C-19 (#8843, the composer attachment menu) was approved in the same direction as the rest of the batch, but never started. It sits before the new cursor, so no future commit range would ever surface it again. The register is the only thing keeping it visible.

On advancing the cursor with C-19 unfinished

The ledger's own rule is that reviewed-through tracks the newest commit for which every candidate has a decision, not an integration. C-19 has a decision. Advancing is correct by that definition — but it is exactly the case where a cursor move can bury real work, which is what DEF-10 exists to prevent.

Reviewed and integrated with Claude Opus 5 in Claude Code.

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

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and adds the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. Records the finding worth carrying forward: upstream landed #8793, #8237 and #8614 within six hours on 2026-08-30, so each later commit carries #8793's composer restructure as context and it was re-litigated in every cherry-pick. It is two independent changes. ComposerSurface was adopted, so Pylon now matches upstream there and those hunks should stop conflicting. The toolbar row is declined permanently, because ComposerToolbarScroller is upstream's own component that they stopped using at four controls and Pylon has sixteen. Register maintenance: DEF-9 retired, adopted as #221. DEF-7 re-checked and still waiting on both premises. DEF-8 re-checked and its blocking condition has cleared - Expo SDK 57 merged as eff44d0 - leaving only its 2026-09-05 floor. Opens DEF-10 for C-19, which is unusual and deliberate: it is approved work that was never started, and it sits before the new cursor, so no future commit range would surface it. The register is the only thing that keeps it visible.

github-actions · 2026-09-01T16:25:35Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.7 KiB	13.6 KiB	−179 B (−1.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.8 KiB	6.6 KiB	−179 B (−2.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.7 KiB	57.2 KiB	−1.5 KiB (−2.5%)	66.4 KiB	✅
Codex	Live turn messages	11	10	−1 (−9.1%)	21	✅
Claude	Total thread wire	13.5 KiB	13.6 KiB	+53 B (+0.4%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−2 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+55 B (+0.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.0 KiB	58.1 KiB	+44 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	9	10	+1 (+11.1%)	21	✅

Baseline: c908bb2 · PR result: ef7340e · 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: 109.5 KiB
Claude decoded thread snapshot: 110.2 KiB

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

rynfar · 2026-09-01T16:18:55Z

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and writes the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. The ledger had been sitting two days behind the integration.

The finding worth carrying forward

Upstream landed #8793 (12:22), #8237 (13:35) and #8614 (18:08) within six hours on 2026-08-30, so every later commit carries #8793's composer restructure in its tree as context. It got re-litigated in every single cherry-pick.

It is two independent changes, and treating them as one decision was the mistake:

ComposerSurface — animated borderRadius, absolute glass layer, bounded collapsed radius. Touches nothing in the toolbar. Adopted, so Pylon's file is now structurally identical to upstream and those hunks should stop conflicting.
The toolbar row — declined permanently. ComposerToolbarScroller is upstream's own component and they still ship it; they stopped using it because their toolbar holds four controls. Pylon's holds sixteen, because ControlPillMenu does not exist upstream at all.

Register maintenance

DEF-9 retired — adopted as fix(web): unify activity logs and composer banners #221, once review established that #8734, not #8693, is the correct base (#8733 restores the pre-#8693 tree exactly, so the revert and re-land cancel).
DEF-7 re-checked, still waiting on both premises. Nothing touched desktop-macos-preview.yml, and its 2026-11-01 floor is unreached.
DEF-8 re-checked, and its blocking condition has cleared: Expo SDK 57 merged as eff44d0cf. Only its 2026-09-05 floor remains, four days out.
DEF-10 opened, and it is deliberately unusual. C-19 (#8843, the composer attachment menu) was approved in the same direction as the rest of the batch, but never started. It sits before the new cursor, so no future commit range would ever surface it again. The register is the only thing keeping it visible.

On advancing the cursor with C-19 unfinished

The ledger's own rule is that reviewed-through tracks the newest commit for which every candidate has a decision, not an integration. C-19 has a decision. Advancing is correct by that definition — but it is exactly the case where a cursor move can bury real work, which is what DEF-10 exists to prevent.

Reviewed and integrated with Claude Opus 5 in Claude Code.

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

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and adds the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. Records the finding worth carrying forward: upstream landed #8793, #8237 and #8614 within six hours on 2026-08-30, so each later commit carries #8793's composer restructure as context and it was re-litigated in every cherry-pick. It is two independent changes. ComposerSurface was adopted, so Pylon now matches upstream there and those hunks should stop conflicting. The toolbar row is declined permanently, because ComposerToolbarScroller is upstream's own component that they stopped using at four controls and Pylon has sixteen. Register maintenance: DEF-9 retired, adopted as #221. DEF-7 re-checked and still waiting on both premises. DEF-8 re-checked and its blocking condition has cleared - Expo SDK 57 merged as eff44d0 - leaving only its 2026-09-05 floor. Opens DEF-10 for C-19, which is unusual and deliberate: it is approved work that was never started, and it sits before the new cursor, so no future commit range would surface it. The register is the only thing that keeps it visible.

github-actions · 2026-09-01T16:25:35Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.7 KiB	13.6 KiB	−179 B (−1.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.8 KiB	6.6 KiB	−179 B (−2.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.7 KiB	57.2 KiB	−1.5 KiB (−2.5%)	66.4 KiB	✅
Codex	Live turn messages	11	10	−1 (−9.1%)	21	✅
Claude	Total thread wire	13.5 KiB	13.6 KiB	+53 B (+0.4%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−2 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+55 B (+0.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.0 KiB	58.1 KiB	+44 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	9	10	+1 (+11.1%)	21	✅

Baseline: c908bb2 · PR result: ef7340e · 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: 109.5 KiB
Claude decoded thread snapshot: 110.2 KiB

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

rynfar · 2026-09-01T16:18:55Z

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and writes the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. The ledger had been sitting two days behind the integration.

The finding worth carrying forward

Upstream landed #8793 (12:22), #8237 (13:35) and #8614 (18:08) within six hours on 2026-08-30, so every later commit carries #8793's composer restructure in its tree as context. It got re-litigated in every single cherry-pick.

It is two independent changes, and treating them as one decision was the mistake:

ComposerSurface — animated borderRadius, absolute glass layer, bounded collapsed radius. Touches nothing in the toolbar. Adopted, so Pylon's file is now structurally identical to upstream and those hunks should stop conflicting.
The toolbar row — declined permanently. ComposerToolbarScroller is upstream's own component and they still ship it; they stopped using it because their toolbar holds four controls. Pylon's holds sixteen, because ControlPillMenu does not exist upstream at all.

Register maintenance

DEF-9 retired — adopted as fix(web): unify activity logs and composer banners #221, once review established that #8734, not #8693, is the correct base (#8733 restores the pre-#8693 tree exactly, so the revert and re-land cancel).
DEF-7 re-checked, still waiting on both premises. Nothing touched desktop-macos-preview.yml, and its 2026-11-01 floor is unreached.
DEF-8 re-checked, and its blocking condition has cleared: Expo SDK 57 merged as eff44d0cf. Only its 2026-09-05 floor remains, four days out.
DEF-10 opened, and it is deliberately unusual. C-19 (#8843, the composer attachment menu) was approved in the same direction as the rest of the batch, but never started. It sits before the new cursor, so no future commit range would ever surface it again. The register is the only thing keeping it visible.

On advancing the cursor with C-19 unfinished

The ledger's own rule is that reviewed-through tracks the newest commit for which every candidate has a decision, not an integration. C-19 has a decision. Advancing is correct by that definition — but it is exactly the case where a cursor move can bury real work, which is what DEF-10 exists to prevent.

Reviewed and integrated with Claude Opus 5 in Claude Code.

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

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and adds the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. Records the finding worth carrying forward: upstream landed #8793, #8237 and #8614 within six hours on 2026-08-30, so each later commit carries #8793's composer restructure as context and it was re-litigated in every cherry-pick. It is two independent changes. ComposerSurface was adopted, so Pylon now matches upstream there and those hunks should stop conflicting. The toolbar row is declined permanently, because ComposerToolbarScroller is upstream's own component that they stopped using at four controls and Pylon has sixteen. Register maintenance: DEF-9 retired, adopted as #221. DEF-7 re-checked and still waiting on both premises. DEF-8 re-checked and its blocking condition has cleared - Expo SDK 57 merged as eff44d0 - leaving only its 2026-09-05 floor. Opens DEF-10 for C-19, which is unusual and deliberate: it is approved work that was never started, and it sits before the new cursor, so no future commit range would surface it. The register is the only thing that keeps it visible.

github-actions · 2026-09-01T16:25:35Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.7 KiB	13.6 KiB	−179 B (−1.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.8 KiB	6.6 KiB	−179 B (−2.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.7 KiB	57.2 KiB	−1.5 KiB (−2.5%)	66.4 KiB	✅
Codex	Live turn messages	11	10	−1 (−9.1%)	21	✅
Claude	Total thread wire	13.5 KiB	13.6 KiB	+53 B (+0.4%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−2 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+55 B (+0.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.0 KiB	58.1 KiB	+44 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	9	10	+1 (+11.1%)	21	✅

Baseline: c908bb2 · PR result: ef7340e · 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: 109.5 KiB
Claude decoded thread snapshot: 110.2 KiB

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

rynfar · 2026-09-01T16:18:55Z

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and writes the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. The ledger had been sitting two days behind the integration.

The finding worth carrying forward

Upstream landed #8793 (12:22), #8237 (13:35) and #8614 (18:08) within six hours on 2026-08-30, so every later commit carries #8793's composer restructure in its tree as context. It got re-litigated in every single cherry-pick.

It is two independent changes, and treating them as one decision was the mistake:

ComposerSurface — animated borderRadius, absolute glass layer, bounded collapsed radius. Touches nothing in the toolbar. Adopted, so Pylon's file is now structurally identical to upstream and those hunks should stop conflicting.
The toolbar row — declined permanently. ComposerToolbarScroller is upstream's own component and they still ship it; they stopped using it because their toolbar holds four controls. Pylon's holds sixteen, because ControlPillMenu does not exist upstream at all.

Register maintenance

DEF-9 retired — adopted as fix(web): unify activity logs and composer banners #221, once review established that #8734, not #8693, is the correct base (#8733 restores the pre-#8693 tree exactly, so the revert and re-land cancel).
DEF-7 re-checked, still waiting on both premises. Nothing touched desktop-macos-preview.yml, and its 2026-11-01 floor is unreached.
DEF-8 re-checked, and its blocking condition has cleared: Expo SDK 57 merged as eff44d0cf. Only its 2026-09-05 floor remains, four days out.
DEF-10 opened, and it is deliberately unusual. C-19 (#8843, the composer attachment menu) was approved in the same direction as the rest of the batch, but never started. It sits before the new cursor, so no future commit range would ever surface it again. The register is the only thing keeping it visible.

On advancing the cursor with C-19 unfinished

The ledger's own rule is that reviewed-through tracks the newest commit for which every candidate has a decision, not an integration. C-19 has a decision. Advancing is correct by that definition — but it is exactly the case where a cursor move can bury real work, which is what DEF-10 exists to prevent.

Reviewed and integrated with Claude Opus 5 in Claude Code.

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

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and adds the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. Records the finding worth carrying forward: upstream landed #8793, #8237 and #8614 within six hours on 2026-08-30, so each later commit carries #8793's composer restructure as context and it was re-litigated in every cherry-pick. It is two independent changes. ComposerSurface was adopted, so Pylon now matches upstream there and those hunks should stop conflicting. The toolbar row is declined permanently, because ComposerToolbarScroller is upstream's own component that they stopped using at four controls and Pylon has sixteen. Register maintenance: DEF-9 retired, adopted as #221. DEF-7 re-checked and still waiting on both premises. DEF-8 re-checked and its blocking condition has cleared - Expo SDK 57 merged as eff44d0 - leaving only its 2026-09-05 floor. Opens DEF-10 for C-19, which is unusual and deliberate: it is approved work that was never started, and it sits before the new cursor, so no future commit range would surface it. The register is the only thing that keeps it visible.

github-actions · 2026-09-01T16:25:35Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.7 KiB	13.6 KiB	−179 B (−1.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.8 KiB	6.6 KiB	−179 B (−2.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.7 KiB	57.2 KiB	−1.5 KiB (−2.5%)	66.4 KiB	✅
Codex	Live turn messages	11	10	−1 (−9.1%)	21	✅
Claude	Total thread wire	13.5 KiB	13.6 KiB	+53 B (+0.4%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−2 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+55 B (+0.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.0 KiB	58.1 KiB	+44 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	9	10	+1 (+11.1%)	21	✅

Baseline: c908bb2 · PR result: ef7340e · 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: 109.5 KiB
Claude decoded thread snapshot: 110.2 KiB

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

rynfar · 2026-09-01T16:18:55Z

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and writes the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. The ledger had been sitting two days behind the integration.

The finding worth carrying forward

Upstream landed #8793 (12:22), #8237 (13:35) and #8614 (18:08) within six hours on 2026-08-30, so every later commit carries #8793's composer restructure in its tree as context. It got re-litigated in every single cherry-pick.

It is two independent changes, and treating them as one decision was the mistake:

ComposerSurface — animated borderRadius, absolute glass layer, bounded collapsed radius. Touches nothing in the toolbar. Adopted, so Pylon's file is now structurally identical to upstream and those hunks should stop conflicting.
The toolbar row — declined permanently. ComposerToolbarScroller is upstream's own component and they still ship it; they stopped using it because their toolbar holds four controls. Pylon's holds sixteen, because ControlPillMenu does not exist upstream at all.

Register maintenance

DEF-9 retired — adopted as fix(web): unify activity logs and composer banners #221, once review established that #8734, not #8693, is the correct base (#8733 restores the pre-#8693 tree exactly, so the revert and re-land cancel).
DEF-7 re-checked, still waiting on both premises. Nothing touched desktop-macos-preview.yml, and its 2026-11-01 floor is unreached.
DEF-8 re-checked, and its blocking condition has cleared: Expo SDK 57 merged as eff44d0cf. Only its 2026-09-05 floor remains, four days out.
DEF-10 opened, and it is deliberately unusual. C-19 (#8843, the composer attachment menu) was approved in the same direction as the rest of the batch, but never started. It sits before the new cursor, so no future commit range would ever surface it again. The register is the only thing keeping it visible.

On advancing the cursor with C-19 unfinished

The ledger's own rule is that reviewed-through tracks the newest commit for which every candidate has a decision, not an integration. C-19 has a decision. Advancing is correct by that definition — but it is exactly the case where a cursor move can bury real work, which is what DEF-10 exists to prevent.

Reviewed and integrated with Claude Opus 5 in Claude Code.

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

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and adds the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. Records the finding worth carrying forward: upstream landed #8793, #8237 and #8614 within six hours on 2026-08-30, so each later commit carries #8793's composer restructure as context and it was re-litigated in every cherry-pick. It is two independent changes. ComposerSurface was adopted, so Pylon now matches upstream there and those hunks should stop conflicting. The toolbar row is declined permanently, because ComposerToolbarScroller is upstream's own component that they stopped using at four controls and Pylon has sixteen. Register maintenance: DEF-9 retired, adopted as #221. DEF-7 re-checked and still waiting on both premises. DEF-8 re-checked and its blocking condition has cleared - Expo SDK 57 merged as eff44d0 - leaving only its 2026-09-05 floor. Opens DEF-10 for C-19, which is unusual and deliberate: it is approved work that was never started, and it sits before the new cursor, so no future commit range would surface it. The register is the only thing that keeps it visible.

github-actions · 2026-09-01T16:25:35Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.7 KiB	13.6 KiB	−179 B (−1.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.8 KiB	6.6 KiB	−179 B (−2.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.7 KiB	57.2 KiB	−1.5 KiB (−2.5%)	66.4 KiB	✅
Codex	Live turn messages	11	10	−1 (−9.1%)	21	✅
Claude	Total thread wire	13.5 KiB	13.6 KiB	+53 B (+0.4%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−2 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+55 B (+0.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.0 KiB	58.1 KiB	+44 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	9	10	+1 (+11.1%)	21	✅

Baseline: c908bb2 · PR result: ef7340e · 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: 109.5 KiB
Claude decoded thread snapshot: 110.2 KiB

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

rynfar · 2026-09-01T16:18:55Z

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and writes the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. The ledger had been sitting two days behind the integration.

The finding worth carrying forward

Upstream landed #8793 (12:22), #8237 (13:35) and #8614 (18:08) within six hours on 2026-08-30, so every later commit carries #8793's composer restructure in its tree as context. It got re-litigated in every single cherry-pick.

It is two independent changes, and treating them as one decision was the mistake:

ComposerSurface — animated borderRadius, absolute glass layer, bounded collapsed radius. Touches nothing in the toolbar. Adopted, so Pylon's file is now structurally identical to upstream and those hunks should stop conflicting.
The toolbar row — declined permanently. ComposerToolbarScroller is upstream's own component and they still ship it; they stopped using it because their toolbar holds four controls. Pylon's holds sixteen, because ControlPillMenu does not exist upstream at all.

Register maintenance

DEF-9 retired — adopted as fix(web): unify activity logs and composer banners #221, once review established that #8734, not #8693, is the correct base (#8733 restores the pre-#8693 tree exactly, so the revert and re-land cancel).
DEF-7 re-checked, still waiting on both premises. Nothing touched desktop-macos-preview.yml, and its 2026-11-01 floor is unreached.
DEF-8 re-checked, and its blocking condition has cleared: Expo SDK 57 merged as eff44d0cf. Only its 2026-09-05 floor remains, four days out.
DEF-10 opened, and it is deliberately unusual. C-19 (#8843, the composer attachment menu) was approved in the same direction as the rest of the batch, but never started. It sits before the new cursor, so no future commit range would ever surface it again. The register is the only thing keeping it visible.

On advancing the cursor with C-19 unfinished

The ledger's own rule is that reviewed-through tracks the newest commit for which every candidate has a decision, not an integration. C-19 has a decision. Advancing is correct by that definition — but it is exactly the case where a cursor move can bury real work, which is what DEF-10 exists to prevent.

Reviewed and integrated with Claude Opus 5 in Claude Code.

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

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and adds the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. Records the finding worth carrying forward: upstream landed #8793, #8237 and #8614 within six hours on 2026-08-30, so each later commit carries #8793's composer restructure as context and it was re-litigated in every cherry-pick. It is two independent changes. ComposerSurface was adopted, so Pylon now matches upstream there and those hunks should stop conflicting. The toolbar row is declined permanently, because ComposerToolbarScroller is upstream's own component that they stopped using at four controls and Pylon has sixteen. Register maintenance: DEF-9 retired, adopted as #221. DEF-7 re-checked and still waiting on both premises. DEF-8 re-checked and its blocking condition has cleared - Expo SDK 57 merged as eff44d0 - leaving only its 2026-09-05 floor. Opens DEF-10 for C-19, which is unusual and deliberate: it is approved work that was never started, and it sits before the new cursor, so no future commit range would surface it. The register is the only thing that keeps it visible.

github-actions · 2026-09-01T16:25:35Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.7 KiB	13.6 KiB	−179 B (−1.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.8 KiB	6.6 KiB	−179 B (−2.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.7 KiB	57.2 KiB	−1.5 KiB (−2.5%)	66.4 KiB	✅
Codex	Live turn messages	11	10	−1 (−9.1%)	21	✅
Claude	Total thread wire	13.5 KiB	13.6 KiB	+53 B (+0.4%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−2 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+55 B (+0.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.0 KiB	58.1 KiB	+44 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	9	10	+1 (+11.1%)	21	✅

Baseline: c908bb2 · PR result: ef7340e · 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: 109.5 KiB
Claude decoded thread snapshot: 110.2 KiB

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

rynfar · 2026-09-01T16:18:55Z

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and writes the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. The ledger had been sitting two days behind the integration.

The finding worth carrying forward

Upstream landed #8793 (12:22), #8237 (13:35) and #8614 (18:08) within six hours on 2026-08-30, so every later commit carries #8793's composer restructure in its tree as context. It got re-litigated in every single cherry-pick.

It is two independent changes, and treating them as one decision was the mistake:

ComposerSurface — animated borderRadius, absolute glass layer, bounded collapsed radius. Touches nothing in the toolbar. Adopted, so Pylon's file is now structurally identical to upstream and those hunks should stop conflicting.
The toolbar row — declined permanently. ComposerToolbarScroller is upstream's own component and they still ship it; they stopped using it because their toolbar holds four controls. Pylon's holds sixteen, because ControlPillMenu does not exist upstream at all.

Register maintenance

DEF-9 retired — adopted as fix(web): unify activity logs and composer banners #221, once review established that #8734, not #8693, is the correct base (#8733 restores the pre-#8693 tree exactly, so the revert and re-land cancel).
DEF-7 re-checked, still waiting on both premises. Nothing touched desktop-macos-preview.yml, and its 2026-11-01 floor is unreached.
DEF-8 re-checked, and its blocking condition has cleared: Expo SDK 57 merged as eff44d0cf. Only its 2026-09-05 floor remains, four days out.
DEF-10 opened, and it is deliberately unusual. C-19 (#8843, the composer attachment menu) was approved in the same direction as the rest of the batch, but never started. It sits before the new cursor, so no future commit range would ever surface it again. The register is the only thing keeping it visible.

On advancing the cursor with C-19 unfinished

The ledger's own rule is that reviewed-through tracks the newest commit for which every candidate has a decision, not an integration. C-19 has a decision. Advancing is correct by that definition — but it is exactly the case where a cursor move can bury real work, which is what DEF-10 exists to prevent.

Reviewed and integrated with Claude Opus 5 in Claude Code.

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

Advances reviewed-through from 1f8ed54ad to 9b2d04317 and adds the batch section for the twenty-nine commits in between, reviewed as C-1 through C-21. Records the finding worth carrying forward: upstream landed #8793, #8237 and #8614 within six hours on 2026-08-30, so each later commit carries #8793's composer restructure as context and it was re-litigated in every cherry-pick. It is two independent changes. ComposerSurface was adopted, so Pylon now matches upstream there and those hunks should stop conflicting. The toolbar row is declined permanently, because ComposerToolbarScroller is upstream's own component that they stopped using at four controls and Pylon has sixteen. Register maintenance: DEF-9 retired, adopted as #221. DEF-7 re-checked and still waiting on both premises. DEF-8 re-checked and its blocking condition has cleared - Expo SDK 57 merged as eff44d0 - leaving only its 2026-09-05 floor. Opens DEF-10 for C-19, which is unusual and deliberate: it is approved work that was never started, and it sits before the new cursor, so no future commit range would surface it. The register is the only thing that keeps it visible.

github-actions · 2026-09-01T16:25:35Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.7 KiB	13.6 KiB	−179 B (−1.3%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.8 KiB	6.6 KiB	−179 B (−2.6%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.7 KiB	57.2 KiB	−1.5 KiB (−2.5%)	66.4 KiB	✅
Codex	Live turn messages	11	10	−1 (−9.1%)	21	✅
Claude	Total thread wire	13.5 KiB	13.6 KiB	+53 B (+0.4%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−2 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+55 B (+0.8%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.0 KiB	58.1 KiB	+44 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	9	10	+1 (+11.1%)	21	✅

Baseline: c908bb2 · PR result: ef7340e · 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: 109.5 KiB
Claude decoded thread snapshot: 110.2 KiB

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

github-actionsBot added vouch:trusted PR author is trusted by repo permissions or the VOUCHED list. size:M labels Sep 1, 2026