fix(web): stop remounting markdown on every activity delta

[juliusmarminge]

Problem

Selecting assistant text on a working thread and clicking "Cite" was flaky: the selection would walk up one block at a time every ~250ms and the toolbar vanished within about a second. Idle threads were fine.

The selection endpoints were being destroyed. MessagesTimeline's sharedState memo built a fresh threadRef via parseScopedThreadKey(routeThreadKey) on every recompute, and it recomputes on every activity delta (agentPanelModel derives from threadActivities). ChatMarkdown keys its react-markdown components map on threadRef, and react-markdown uses those functions as React element types, so a new identity remounted every <p>/<pre>/<code> in every visible assistant message. Chrome repairs a selection whose nodes were removed to the nearest boundary, hence the one-block-per-tick walk.

This predates the cite feature; it was already re-creating shiki blocks and chips on every delta.

Fix

Reuse the already-memoized citationThreadRef so ChatMarkdown's props stay referentially stable and its memo bails out on activity ticks.

Verified with vp lint, apps/web typecheck, and the MessagesTimeline tests (114 passing).

---

Claude Fable 5 via Claude Code

🤖 Generated with Claude Code

---

Note

Low Risk
Single-line behavioral fix reusing an existing memoized value; no API or security changes, with tests already noted as passing.

Overview
Fixes flaky assistant text selection and Cite on working threads by keeping threadRef referentially stable in timeline row shared context.

MessagesTimeline no longer calls parseScopedThreadKey(routeThreadKey) inside the sharedState memo (which recomputes on each activity tick via agentPanelModel). It now passes the existing memoized citationThreadRef into context so ChatMarkdown props stay stable, react-markdown does not treat new component types on every delta, and assistant markdown blocks are not remounted while the agent is active.

^{Reviewed by Cursor Bugbot for commit ff3883a. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Stop remounting markdown in MessagesTimeline on activity delta

The shared timeline state now uses the existing citationThreadRef instead of building a new object from parseScopedThreadKey(routeThreadKey). citationThreadRef is added to the useMemo dependency list so its stable reference carries across activity-driven recomputations. This keeps ChatMarkdown's component map from being recreated on every update.

^{Macroscope summarized ff3883a.}

[github-actions]

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	13.4 KiB	+161 B (+1.2%)	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	6.4 KiB	6.6 KiB	+161 B (+2.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.6 KiB	57.0 KiB	+1.4 KiB (+2.6%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+10 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−3 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+13 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.8 KiB	57.8 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	10	10	0 (0.0%)	21	✅

Baseline: ef6cc0b · PR result: ff3883a · 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: 110.1 KiB

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

[macroscopeapp]

Approvability

Verdict: Approved at ff3883a

Macroscope's review found this PR approvable — This is a narrowly scoped rendering fix that preserves the existing thread identity while preventing markdown components from remounting during activity updates. It does not introduce new capabilities, schema changes, configuration changes, or sensitive-path modifications.

^{You can add or adjust custom eligibility rules. Learn more.}

[juliusmarminge]

Problem

Selecting assistant text on a working thread and clicking "Cite" was flaky: the selection would walk up one block at a time every ~250ms and the toolbar vanished within about a second. Idle threads were fine.

The selection endpoints were being destroyed. MessagesTimeline's sharedState memo built a fresh threadRef via parseScopedThreadKey(routeThreadKey) on every recompute, and it recomputes on every activity delta (agentPanelModel derives from threadActivities). ChatMarkdown keys its react-markdown components map on threadRef, and react-markdown uses those functions as React element types, so a new identity remounted every <p>/<pre>/<code> in every visible assistant message. Chrome repairs a selection whose nodes were removed to the nearest boundary, hence the one-block-per-tick walk.

This predates the cite feature; it was already re-creating shiki blocks and chips on every delta.

Fix

Reuse the already-memoized citationThreadRef so ChatMarkdown's props stay referentially stable and its memo bails out on activity ticks.

Verified with vp lint, apps/web typecheck, and the MessagesTimeline tests (114 passing).

---

Claude Fable 5 via Claude Code

🤖 Generated with Claude Code

---

Note

Low Risk
Single-line behavioral fix reusing an existing memoized value; no API or security changes, with tests already noted as passing.

Overview
Fixes flaky assistant text selection and Cite on working threads by keeping threadRef referentially stable in timeline row shared context.

MessagesTimeline no longer calls parseScopedThreadKey(routeThreadKey) inside the sharedState memo (which recomputes on each activity tick via agentPanelModel). It now passes the existing memoized citationThreadRef into context so ChatMarkdown props stay stable, react-markdown does not treat new component types on every delta, and assistant markdown blocks are not remounted while the agent is active.

^{Reviewed by Cursor Bugbot for commit ff3883a. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Stop remounting markdown in MessagesTimeline on activity delta

The shared timeline state now uses the existing citationThreadRef instead of building a new object from parseScopedThreadKey(routeThreadKey). citationThreadRef is added to the useMemo dependency list so its stable reference carries across activity-driven recomputations. This keeps ChatMarkdown's component map from being recreated on every update.

^{Macroscope summarized ff3883a.}

[github-actions]

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	13.4 KiB	+161 B (+1.2%)	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	6.4 KiB	6.6 KiB	+161 B (+2.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.6 KiB	57.0 KiB	+1.4 KiB (+2.6%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+10 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−3 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+13 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.8 KiB	57.8 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	10	10	0 (0.0%)	21	✅

Baseline: ef6cc0b · PR result: ff3883a · 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: 110.1 KiB

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

[macroscopeapp]

Approvability

Verdict: Approved at ff3883a

Macroscope's review found this PR approvable — This is a narrowly scoped rendering fix that preserves the existing thread identity while preventing markdown components from remounting during activity updates. It does not introduce new capabilities, schema changes, configuration changes, or sensitive-path modifications.

^{You can add or adjust custom eligibility rules. Learn more.}

[juliusmarminge]

Problem

Selecting assistant text on a working thread and clicking "Cite" was flaky: the selection would walk up one block at a time every ~250ms and the toolbar vanished within about a second. Idle threads were fine.

The selection endpoints were being destroyed. MessagesTimeline's sharedState memo built a fresh threadRef via parseScopedThreadKey(routeThreadKey) on every recompute, and it recomputes on every activity delta (agentPanelModel derives from threadActivities). ChatMarkdown keys its react-markdown components map on threadRef, and react-markdown uses those functions as React element types, so a new identity remounted every <p>/<pre>/<code> in every visible assistant message. Chrome repairs a selection whose nodes were removed to the nearest boundary, hence the one-block-per-tick walk.

This predates the cite feature; it was already re-creating shiki blocks and chips on every delta.

Fix

Reuse the already-memoized citationThreadRef so ChatMarkdown's props stay referentially stable and its memo bails out on activity ticks.

Verified with vp lint, apps/web typecheck, and the MessagesTimeline tests (114 passing).

---

Claude Fable 5 via Claude Code

🤖 Generated with Claude Code

---

Note

Low Risk
Single-line behavioral fix reusing an existing memoized value; no API or security changes, with tests already noted as passing.

Overview
Fixes flaky assistant text selection and Cite on working threads by keeping threadRef referentially stable in timeline row shared context.

MessagesTimeline no longer calls parseScopedThreadKey(routeThreadKey) inside the sharedState memo (which recomputes on each activity tick via agentPanelModel). It now passes the existing memoized citationThreadRef into context so ChatMarkdown props stay stable, react-markdown does not treat new component types on every delta, and assistant markdown blocks are not remounted while the agent is active.

^{Reviewed by Cursor Bugbot for commit ff3883a. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Stop remounting markdown in MessagesTimeline on activity delta

The shared timeline state now uses the existing citationThreadRef instead of building a new object from parseScopedThreadKey(routeThreadKey). citationThreadRef is added to the useMemo dependency list so its stable reference carries across activity-driven recomputations. This keeps ChatMarkdown's component map from being recreated on every update.

^{Macroscope summarized ff3883a.}

[github-actions]

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	13.4 KiB	+161 B (+1.2%)	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	6.4 KiB	6.6 KiB	+161 B (+2.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.6 KiB	57.0 KiB	+1.4 KiB (+2.6%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+10 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−3 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+13 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.8 KiB	57.8 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	10	10	0 (0.0%)	21	✅

Baseline: ef6cc0b · PR result: ff3883a · 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: 110.1 KiB

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

[macroscopeapp]

Approvability

Verdict: Approved at ff3883a

Macroscope's review found this PR approvable — This is a narrowly scoped rendering fix that preserves the existing thread identity while preventing markdown components from remounting during activity updates. It does not introduce new capabilities, schema changes, configuration changes, or sensitive-path modifications.

^{You can add or adjust custom eligibility rules. Learn more.}

[juliusmarminge]

Problem

Selecting assistant text on a working thread and clicking "Cite" was flaky: the selection would walk up one block at a time every ~250ms and the toolbar vanished within about a second. Idle threads were fine.

The selection endpoints were being destroyed. MessagesTimeline's sharedState memo built a fresh threadRef via parseScopedThreadKey(routeThreadKey) on every recompute, and it recomputes on every activity delta (agentPanelModel derives from threadActivities). ChatMarkdown keys its react-markdown components map on threadRef, and react-markdown uses those functions as React element types, so a new identity remounted every <p>/<pre>/<code> in every visible assistant message. Chrome repairs a selection whose nodes were removed to the nearest boundary, hence the one-block-per-tick walk.

This predates the cite feature; it was already re-creating shiki blocks and chips on every delta.

Fix

Reuse the already-memoized citationThreadRef so ChatMarkdown's props stay referentially stable and its memo bails out on activity ticks.

Verified with vp lint, apps/web typecheck, and the MessagesTimeline tests (114 passing).

---

Claude Fable 5 via Claude Code

🤖 Generated with Claude Code

---

Note

Low Risk
Single-line behavioral fix reusing an existing memoized value; no API or security changes, with tests already noted as passing.

Overview
Fixes flaky assistant text selection and Cite on working threads by keeping threadRef referentially stable in timeline row shared context.

MessagesTimeline no longer calls parseScopedThreadKey(routeThreadKey) inside the sharedState memo (which recomputes on each activity tick via agentPanelModel). It now passes the existing memoized citationThreadRef into context so ChatMarkdown props stay stable, react-markdown does not treat new component types on every delta, and assistant markdown blocks are not remounted while the agent is active.

^{Reviewed by Cursor Bugbot for commit ff3883a. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Stop remounting markdown in MessagesTimeline on activity delta

The shared timeline state now uses the existing citationThreadRef instead of building a new object from parseScopedThreadKey(routeThreadKey). citationThreadRef is added to the useMemo dependency list so its stable reference carries across activity-driven recomputations. This keeps ChatMarkdown's component map from being recreated on every update.

^{Macroscope summarized ff3883a.}

[github-actions]

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	13.4 KiB	+161 B (+1.2%)	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	6.4 KiB	6.6 KiB	+161 B (+2.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.6 KiB	57.0 KiB	+1.4 KiB (+2.6%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+10 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−3 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+13 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.8 KiB	57.8 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	10	10	0 (0.0%)	21	✅

Baseline: ef6cc0b · PR result: ff3883a · 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: 110.1 KiB

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

[macroscopeapp]

Approvability

Verdict: Approved at ff3883a

Macroscope's review found this PR approvable — This is a narrowly scoped rendering fix that preserves the existing thread identity while preventing markdown components from remounting during activity updates. It does not introduce new capabilities, schema changes, configuration changes, or sensitive-path modifications.

^{You can add or adjust custom eligibility rules. Learn more.}

[juliusmarminge]

Problem

Selecting assistant text on a working thread and clicking "Cite" was flaky: the selection would walk up one block at a time every ~250ms and the toolbar vanished within about a second. Idle threads were fine.

The selection endpoints were being destroyed. MessagesTimeline's sharedState memo built a fresh threadRef via parseScopedThreadKey(routeThreadKey) on every recompute, and it recomputes on every activity delta (agentPanelModel derives from threadActivities). ChatMarkdown keys its react-markdown components map on threadRef, and react-markdown uses those functions as React element types, so a new identity remounted every <p>/<pre>/<code> in every visible assistant message. Chrome repairs a selection whose nodes were removed to the nearest boundary, hence the one-block-per-tick walk.

This predates the cite feature; it was already re-creating shiki blocks and chips on every delta.

Fix

Reuse the already-memoized citationThreadRef so ChatMarkdown's props stay referentially stable and its memo bails out on activity ticks.

Verified with vp lint, apps/web typecheck, and the MessagesTimeline tests (114 passing).

---

Claude Fable 5 via Claude Code

🤖 Generated with Claude Code

---

Note

Low Risk
Single-line behavioral fix reusing an existing memoized value; no API or security changes, with tests already noted as passing.

Overview
Fixes flaky assistant text selection and Cite on working threads by keeping threadRef referentially stable in timeline row shared context.

MessagesTimeline no longer calls parseScopedThreadKey(routeThreadKey) inside the sharedState memo (which recomputes on each activity tick via agentPanelModel). It now passes the existing memoized citationThreadRef into context so ChatMarkdown props stay stable, react-markdown does not treat new component types on every delta, and assistant markdown blocks are not remounted while the agent is active.

^{Reviewed by Cursor Bugbot for commit ff3883a. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Stop remounting markdown in MessagesTimeline on activity delta

The shared timeline state now uses the existing citationThreadRef instead of building a new object from parseScopedThreadKey(routeThreadKey). citationThreadRef is added to the useMemo dependency list so its stable reference carries across activity-driven recomputations. This keeps ChatMarkdown's component map from being recreated on every update.

^{Macroscope summarized ff3883a.}

[github-actions]

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	13.4 KiB	+161 B (+1.2%)	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	6.4 KiB	6.6 KiB	+161 B (+2.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.6 KiB	57.0 KiB	+1.4 KiB (+2.6%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+10 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−3 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+13 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.8 KiB	57.8 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	10	10	0 (0.0%)	21	✅

Baseline: ef6cc0b · PR result: ff3883a · 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: 110.1 KiB

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

[macroscopeapp]

Approvability

Verdict: Approved at ff3883a

Macroscope's review found this PR approvable — This is a narrowly scoped rendering fix that preserves the existing thread identity while preventing markdown components from remounting during activity updates. It does not introduce new capabilities, schema changes, configuration changes, or sensitive-path modifications.

^{You can add or adjust custom eligibility rules. Learn more.}

[juliusmarminge]

Problem

Selecting assistant text on a working thread and clicking "Cite" was flaky: the selection would walk up one block at a time every ~250ms and the toolbar vanished within about a second. Idle threads were fine.

The selection endpoints were being destroyed. MessagesTimeline's sharedState memo built a fresh threadRef via parseScopedThreadKey(routeThreadKey) on every recompute, and it recomputes on every activity delta (agentPanelModel derives from threadActivities). ChatMarkdown keys its react-markdown components map on threadRef, and react-markdown uses those functions as React element types, so a new identity remounted every <p>/<pre>/<code> in every visible assistant message. Chrome repairs a selection whose nodes were removed to the nearest boundary, hence the one-block-per-tick walk.

This predates the cite feature; it was already re-creating shiki blocks and chips on every delta.

Fix

Reuse the already-memoized citationThreadRef so ChatMarkdown's props stay referentially stable and its memo bails out on activity ticks.

Verified with vp lint, apps/web typecheck, and the MessagesTimeline tests (114 passing).

---

Claude Fable 5 via Claude Code

🤖 Generated with Claude Code

---

Note

Low Risk
Single-line behavioral fix reusing an existing memoized value; no API or security changes, with tests already noted as passing.

Overview
Fixes flaky assistant text selection and Cite on working threads by keeping threadRef referentially stable in timeline row shared context.

MessagesTimeline no longer calls parseScopedThreadKey(routeThreadKey) inside the sharedState memo (which recomputes on each activity tick via agentPanelModel). It now passes the existing memoized citationThreadRef into context so ChatMarkdown props stay stable, react-markdown does not treat new component types on every delta, and assistant markdown blocks are not remounted while the agent is active.

^{Reviewed by Cursor Bugbot for commit ff3883a. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Stop remounting markdown in MessagesTimeline on activity delta

The shared timeline state now uses the existing citationThreadRef instead of building a new object from parseScopedThreadKey(routeThreadKey). citationThreadRef is added to the useMemo dependency list so its stable reference carries across activity-driven recomputations. This keeps ChatMarkdown's component map from being recreated on every update.

^{Macroscope summarized ff3883a.}

[github-actions]

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	13.4 KiB	+161 B (+1.2%)	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	6.4 KiB	6.6 KiB	+161 B (+2.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.6 KiB	57.0 KiB	+1.4 KiB (+2.6%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+10 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−3 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+13 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.8 KiB	57.8 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	10	10	0 (0.0%)	21	✅

Baseline: ef6cc0b · PR result: ff3883a · 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: 110.1 KiB

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

[macroscopeapp]

Approvability

Verdict: Approved at ff3883a

Macroscope's review found this PR approvable — This is a narrowly scoped rendering fix that preserves the existing thread identity while preventing markdown components from remounting during activity updates. It does not introduce new capabilities, schema changes, configuration changes, or sensitive-path modifications.

^{You can add or adjust custom eligibility rules. Learn more.}

[juliusmarminge]

Problem

Selecting assistant text on a working thread and clicking "Cite" was flaky: the selection would walk up one block at a time every ~250ms and the toolbar vanished within about a second. Idle threads were fine.

The selection endpoints were being destroyed. MessagesTimeline's sharedState memo built a fresh threadRef via parseScopedThreadKey(routeThreadKey) on every recompute, and it recomputes on every activity delta (agentPanelModel derives from threadActivities). ChatMarkdown keys its react-markdown components map on threadRef, and react-markdown uses those functions as React element types, so a new identity remounted every <p>/<pre>/<code> in every visible assistant message. Chrome repairs a selection whose nodes were removed to the nearest boundary, hence the one-block-per-tick walk.

This predates the cite feature; it was already re-creating shiki blocks and chips on every delta.

Fix

Reuse the already-memoized citationThreadRef so ChatMarkdown's props stay referentially stable and its memo bails out on activity ticks.

Verified with vp lint, apps/web typecheck, and the MessagesTimeline tests (114 passing).

---

Claude Fable 5 via Claude Code

🤖 Generated with Claude Code

---

Note

Low Risk
Single-line behavioral fix reusing an existing memoized value; no API or security changes, with tests already noted as passing.

Overview
Fixes flaky assistant text selection and Cite on working threads by keeping threadRef referentially stable in timeline row shared context.

MessagesTimeline no longer calls parseScopedThreadKey(routeThreadKey) inside the sharedState memo (which recomputes on each activity tick via agentPanelModel). It now passes the existing memoized citationThreadRef into context so ChatMarkdown props stay stable, react-markdown does not treat new component types on every delta, and assistant markdown blocks are not remounted while the agent is active.

^{Reviewed by Cursor Bugbot for commit ff3883a. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Stop remounting markdown in MessagesTimeline on activity delta

The shared timeline state now uses the existing citationThreadRef instead of building a new object from parseScopedThreadKey(routeThreadKey). citationThreadRef is added to the useMemo dependency list so its stable reference carries across activity-driven recomputations. This keeps ChatMarkdown's component map from being recreated on every update.

^{Macroscope summarized ff3883a.}

[github-actions]

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	13.4 KiB	+161 B (+1.2%)	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	6.4 KiB	6.6 KiB	+161 B (+2.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.6 KiB	57.0 KiB	+1.4 KiB (+2.6%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+10 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−3 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+13 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.8 KiB	57.8 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	10	10	0 (0.0%)	21	✅

Baseline: ef6cc0b · PR result: ff3883a · 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: 110.1 KiB

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

[macroscopeapp]

Approvability

Verdict: Approved at ff3883a

Macroscope's review found this PR approvable — This is a narrowly scoped rendering fix that preserves the existing thread identity while preventing markdown components from remounting during activity updates. It does not introduce new capabilities, schema changes, configuration changes, or sensitive-path modifications.

^{You can add or adjust custom eligibility rules. Learn more.}

[juliusmarminge]

Problem

Selecting assistant text on a working thread and clicking "Cite" was flaky: the selection would walk up one block at a time every ~250ms and the toolbar vanished within about a second. Idle threads were fine.

The selection endpoints were being destroyed. MessagesTimeline's sharedState memo built a fresh threadRef via parseScopedThreadKey(routeThreadKey) on every recompute, and it recomputes on every activity delta (agentPanelModel derives from threadActivities). ChatMarkdown keys its react-markdown components map on threadRef, and react-markdown uses those functions as React element types, so a new identity remounted every <p>/<pre>/<code> in every visible assistant message. Chrome repairs a selection whose nodes were removed to the nearest boundary, hence the one-block-per-tick walk.

This predates the cite feature; it was already re-creating shiki blocks and chips on every delta.

Fix

Reuse the already-memoized citationThreadRef so ChatMarkdown's props stay referentially stable and its memo bails out on activity ticks.

Verified with vp lint, apps/web typecheck, and the MessagesTimeline tests (114 passing).

---

Claude Fable 5 via Claude Code

🤖 Generated with Claude Code

---

Note

Low Risk
Single-line behavioral fix reusing an existing memoized value; no API or security changes, with tests already noted as passing.

Overview
Fixes flaky assistant text selection and Cite on working threads by keeping threadRef referentially stable in timeline row shared context.

MessagesTimeline no longer calls parseScopedThreadKey(routeThreadKey) inside the sharedState memo (which recomputes on each activity tick via agentPanelModel). It now passes the existing memoized citationThreadRef into context so ChatMarkdown props stay stable, react-markdown does not treat new component types on every delta, and assistant markdown blocks are not remounted while the agent is active.

^{Reviewed by Cursor Bugbot for commit ff3883a. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Stop remounting markdown in MessagesTimeline on activity delta

The shared timeline state now uses the existing citationThreadRef instead of building a new object from parseScopedThreadKey(routeThreadKey). citationThreadRef is added to the useMemo dependency list so its stable reference carries across activity-driven recomputations. This keeps ChatMarkdown's component map from being recreated on every update.

^{Macroscope summarized ff3883a.}

[github-actions]

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	13.4 KiB	+161 B (+1.2%)	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	6.4 KiB	6.6 KiB	+161 B (+2.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.6 KiB	57.0 KiB	+1.4 KiB (+2.6%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+10 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−3 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	+13 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.8 KiB	57.8 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	10	10	0 (0.0%)	21	✅

Baseline: ef6cc0b · PR result: ff3883a · 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: 110.1 KiB

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

[macroscopeapp]

Approvability

Verdict: Approved at ff3883a

Macroscope's review found this PR approvable — This is a narrowly scoped rendering fix that preserves the existing thread identity while preventing markdown components from remounting during activity updates. It does not introduce new capabilities, schema changes, configuration changes, or sensitive-path modifications.

^{You can add or adjust custom eligibility rules. Learn more.}