Repository navigation
docs(marketing): performance changelog for 2026-07-19 - #5781
Conversation
|
Ready to review this PR? Stage has broken it down into 4 individual chapters for you:
Chapters generated by Stage for commit 437117e on Jul 19, 2026 7:21pm UTC. |
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughAdds July 19 performance and terminal release content, matching tweet entries, and draft-aware changelog loading that excludes draft pages from rendered entries and slugs. ChangesChangelog publications and filtering
Estimated code review effort: 2 (Simple) | ~10 minutes 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (3)
apps/marketing/content/changelog/2026-07-19-performance-memory-and-load.mdx (2)
10-10: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueUse a more concise synonym.
Consider replacing "lots of" with "many" for a more professional tone, as suggested by static analysis. As per path instructions, ensure you run
bun run lint:fixand verifybun run lintexits successfully after applying your edits.💡 Proposed style fix
-Every hidden terminal used to hold its full scrollback and a live GPU context forever, so a long session with lots of background terminals could grow to well over a gigabyte of memory. Superset now keeps only your most-recently-used terminals fully in memory and rebuilds the rest from disk when you return to them. The underlying process keeps running the whole time. +Every hidden terminal used to hold its full scrollback and a live GPU context forever, so a long session with many background terminals could grow to well over a gigabyte of memory. Superset now keeps only your most-recently-used terminals fully in memory and rebuilds the rest from disk when you return to them. The underlying process keeps running the whole time.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@apps/marketing/content/changelog/2026-07-19-performance-memory-and-load.mdx` at line 10, In the changelog sentence, replace the phrase “lots of background terminals” with the more concise “many background terminals.” Run bun run lint:fix, then verify bun run lint completes successfully.Sources: Path instructions, Linters/SAST tools
1-4: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueConsider adding a description to the frontmatter for SEO.
The frontmatter parser explicitly supports an optional
descriptionfield. Adding a brief summary improves metadata and search result snippets for this changelog entry.💡 Proposed frontmatter
--- title: "Performance: lighter memory and smoother under heavy load" +description: "Performance improvements: lower memory usage for terminals, smoother git status, and lighter background port detection." date: 2026-07-19 ---🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@apps/marketing/content/changelog/2026-07-19-performance-memory-and-load.mdx` around lines 1 - 4, Add a concise SEO-oriented description field to the frontmatter of the changelog entry, summarizing its performance, memory, and heavy-load improvements while preserving the existing title and date metadata.apps/marketing/content/changelog/2026-07-19-performance-tweet.md (1)
9-9: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueUse a more concise synonym.
Consider replacing "a lot of" with "many" to be more concise, as suggested by static analysis. As per path instructions, ensure you run
bun run lint:fixand verifybun run lintexits successfully after editing.💡 Proposed style fix
-Every hidden terminal used to keep its full scrollback and a GPU context alive forever, so a long session with a lot of background terminals could climb past a gigabyte of memory. Now we keep your recent terminals in memory and rebuild the rest from disk when you come back to them. The process keeps running the whole time. In a 16-terminal test that cut renderer memory from 1,127 MB to 838 MB and JS heap from 581 MB to 286 MB. Tune it in Settings → Terminal → Background terminal memory. +Every hidden terminal used to keep its full scrollback and a GPU context alive forever, so a long session with many background terminals could climb past a gigabyte of memory. Now we keep your recent terminals in memory and rebuild the rest from disk when you come back to them. The process keeps running the whole time. In a 16-terminal test that cut renderer memory from 1,127 MB to 838 MB and JS heap from 581 MB to 286 MB. Tune it in Settings → Terminal → Background terminal memory.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@apps/marketing/content/changelog/2026-07-19-performance-tweet.md` at line 9, In the changelog paragraph, replace the phrase “a lot of” with the more concise synonym “many” while preserving the surrounding meaning. After editing, run bun run lint:fix and verify that bun run lint exits successfully.Sources: Path instructions, Linters/SAST tools
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@apps/marketing/content/changelog/2026-07-19-performance-tweet.md`:
- Line 13: Update the performance comparison sentence near the “6 scan cycles”
reference to replace “over the same window” with an explicit timeframe of 6 scan
cycles, ensuring the reduction from 90 to 6 ps spawns is understandable
standalone.
In
`@apps/marketing/content/changelog/2026-07-19-terminal-restore-failed-agents.mdx`:
- Line 6: Add the missing hero image and inline screenshots referenced by the
TODO comment in the changelog, covering background-terminal restore, the red
“failed” agent indicator, and ⌘F in the Changes pane; remove the TODO once the
assets are included.
---
Nitpick comments:
In `@apps/marketing/content/changelog/2026-07-19-performance-memory-and-load.mdx`:
- Line 10: In the changelog sentence, replace the phrase “lots of background
terminals” with the more concise “many background terminals.” Run bun run
lint:fix, then verify bun run lint completes successfully.
- Around line 1-4: Add a concise SEO-oriented description field to the
frontmatter of the changelog entry, summarizing its performance, memory, and
heavy-load improvements while preserving the existing title and date metadata.
In `@apps/marketing/content/changelog/2026-07-19-performance-tweet.md`:
- Line 9: In the changelog paragraph, replace the phrase “a lot of” with the
more concise synonym “many” while preserving the surrounding meaning. After
editing, run bun run lint:fix and verify that bun run lint exits successfully.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 23362b6f-f39e-4ab3-877c-2af2b08bed7e
⛔ Files ignored due to path filters (6)
apps/marketing/public/changelog/2026-07-19-git-typical-stall-ab.pngis excluded by!**/*.pngapps/marketing/public/changelog/2026-07-19-git-worker-pool-ab.pngis excluded by!**/*.pngapps/marketing/public/changelog/2026-07-19-port-idle-cadence-ab.pngis excluded by!**/*.pngapps/marketing/public/changelog/2026-07-19-port-scanner-ab.pngis excluded by!**/*.pngapps/marketing/public/changelog/2026-07-19-terminal-gpu-contexts-ab.pngis excluded by!**/*.pngapps/marketing/public/changelog/2026-07-19-terminal-memory-ab.pngis excluded by!**/*.png
📒 Files selected for processing (4)
apps/marketing/content/changelog/2026-07-19-performance-memory-and-load.mdxapps/marketing/content/changelog/2026-07-19-performance-tweet.mdapps/marketing/content/changelog/2026-07-19-terminal-restore-failed-agents.mdxapps/marketing/content/changelog/2026-07-19-tweet.md
|
|
||
| Git status in big repos used to run on the same thread that relays your terminal output, so a branch switch or a burst of file changes could make the app hitch. It runs on a background worker pool now. Profiling 8 workspaces churning a 20k-file repo, the worst-case stall on that thread dropped from 82 ms to 28 ms. | ||
|
|
||
| We also cut background CPU from port detection: one shared process scan per cycle instead of one per terminal. With 15 terminals open that dropped background ps spawns from 90 to 6 over the same window, and the gap grows the more terminals you have. |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win
Clarify the missing time window reference.
The phrase "over the same window" lacks an antecedent in the tweet text. The changelog explicitly mentions "6 scan cycles" earlier, but this context is omitted here, making the metric unclear. State the timeframe explicitly so the statistic stands alone cleanly.
💚 Proposed fix to explicitly state the window
-We also cut background CPU from port detection: one shared process scan per cycle instead of one per terminal. With 15 terminals open that dropped background ps spawns from 90 to 6 over the same window, and the gap grows the more terminals you have.
+We also cut background CPU from port detection: one shared process scan per cycle instead of one per terminal. With 15 terminals open that dropped background ps spawns from 90 to 6 over 6 cycles, and the gap grows the more terminals you have.📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| We also cut background CPU from port detection: one shared process scan per cycle instead of one per terminal. With 15 terminals open that dropped background ps spawns from 90 to 6 over the same window, and the gap grows the more terminals you have. | |
| We also cut background CPU from port detection: one shared process scan per cycle instead of one per terminal. With 15 terminals open that dropped background ps spawns from 90 to 6 over 6 cycles, and the gap grows the more terminals you have. |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@apps/marketing/content/changelog/2026-07-19-performance-tweet.md` at line 13,
Update the performance comparison sentence near the “6 scan cycles” reference to
replace “over the same window” with an explicit timeframe of 6 scan cycles,
ensuring the reduction from 90 to 6 ps spawns is understandable standalone.
| date: 2026-07-19 | ||
| --- | ||
|
|
||
| {/* TODO: hero + inline shots — capture on a machine that can launch the desktop app (background-terminal restore, red "failed" agent dot, ⌘F in the Changes pane) */} |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win
Provide the missing hero and inline shots.
The TODO comment indicates that visual assets (hero image, background-terminal restore, red "failed" agent dot, and Changes pane search) still need to be captured and added. Ensure these are included before final publication so readers have the intended visual context.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In
`@apps/marketing/content/changelog/2026-07-19-terminal-restore-failed-agents.mdx`
at line 6, Add the missing hero image and inline screenshots referenced by the
TODO comment in the changelog, covering background-terminal restore, the red
“failed” agent indicator, and ⌘F in the Changes pane; remove the TODO once the
assets are included.
Greptile SummaryThis PR adds two dated changelog entries for 2026-07-19 — a terminal-reliability weekly changelog and a performance-focused post with six before/after benchmark charts — along with their companion tweet drafts and the corresponding PNG assets.
Confidence Score: 4/5Docs/content-only change; no code paths are affected. The weekly tweet draft contains a stray meta-sentence as plain body text that would appear publicly if the file is published without a manual edit pass. The performance post and its tweet are clean — numbers are internally consistent, all image paths resolve, and the writing is ready to publish. The weekly tweet draft has an automation-attribution sentence sitting as plain body text between content paragraphs, which would surface publicly if the file is consumed as-is. The weekly MDX has an unresolved TODO calling out missing hero and inline screenshots, leaving that post image-free compared to its performance sibling.
|
| Filename | Overview |
|---|---|
| apps/marketing/content/changelog/2026-07-19-tweet.md | Weekly tweet draft — content is accurate and consistent with the MDX changelog, but contains a literal "This changelog draft was created by an automation." sentence that would appear as public tweet text if published as-is. |
| apps/marketing/content/changelog/2026-07-19-terminal-restore-failed-agents.mdx | Weekly changelog entry — well-structured with PRBadge links and detailed feature/fix list, but has an unresolved TODO comment flagging missing hero and inline screenshots that were planned. |
| apps/marketing/content/changelog/2026-07-19-performance-memory-and-load.mdx | Performance changelog entry — numbers are internally consistent across tables, alt text, and prose; all 6 image references resolve to files added in this PR. |
| apps/marketing/content/changelog/2026-07-19-performance-tweet.md | Performance tweet draft — clean and consistent with the performance MDX; numbers match and no stray metadata text. |
| apps/marketing/public/changelog/2026-07-19-terminal-memory-ab.png | Binary image asset; referenced correctly in the performance MDX. |
| apps/marketing/public/changelog/2026-07-19-terminal-gpu-contexts-ab.png | Binary image asset; referenced correctly in the performance MDX. |
| apps/marketing/public/changelog/2026-07-19-git-worker-pool-ab.png | Binary image asset; referenced correctly in the performance MDX. |
| apps/marketing/public/changelog/2026-07-19-git-typical-stall-ab.png | Binary image asset; referenced correctly in the performance MDX. |
| apps/marketing/public/changelog/2026-07-19-port-scanner-ab.png | Binary image asset; referenced correctly in the performance MDX. |
| apps/marketing/public/changelog/2026-07-19-port-idle-cadence-ab.png | Binary image asset; referenced correctly in the performance MDX. |
Flowchart
%%{init: {'theme': 'neutral'}}%%
flowchart TD
subgraph Weekly["Weekly Changelog"]
W1["Running terminals restored on open (#5740)"]
W2["Failed-agent red status (#5697)"]
W3["Find in Changes pane (#5648)"]
W4["Confirm before closing terminal (#5517)"]
W_TODO["⚠️ TODO: hero + inline screenshots missing"]
end
subgraph Perf["Performance Changelog"]
P1["Memory eviction: RSS 1127 to 838 MB"]
P2["Git worker pool: stall 82 to 28 ms"]
P3["Port scanner: ps spawns 15 to 1 per cycle"]
P_CHARTS["6 before/after PNG charts added"]
end
subgraph Tweets["Tweet Drafts"]
TW["Weekly tweet (2026-07-19-tweet.md)"]
TP["Performance tweet (clean)"]
TW_WARN["⚠️ Stray automation attribution in body text"]
end
Weekly --> TW
Perf --> TP
TW --> TW_WARN
%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
flowchart TD
subgraph Weekly["Weekly Changelog"]
W1["Running terminals restored on open (#5740)"]
W2["Failed-agent red status (#5697)"]
W3["Find in Changes pane (#5648)"]
W4["Confirm before closing terminal (#5517)"]
W_TODO["⚠️ TODO: hero + inline screenshots missing"]
end
subgraph Perf["Performance Changelog"]
P1["Memory eviction: RSS 1127 to 838 MB"]
P2["Git worker pool: stall 82 to 28 ms"]
P3["Port scanner: ps spawns 15 to 1 per cycle"]
P_CHARTS["6 before/after PNG charts added"]
end
subgraph Tweets["Tweet Drafts"]
TW["Weekly tweet (2026-07-19-tweet.md)"]
TP["Performance tweet (clean)"]
TW_WARN["⚠️ Stray automation attribution in body text"]
end
Weekly --> TW
Perf --> TP
TW --> TW_WARN
Prompt To Fix All With AI
Fix the following 2 code review issues. Work through them one at a time, proposing concise fixes.
---
### Issue 1 of 2
apps/marketing/content/changelog/2026-07-19-tweet.md:23
**Automation attribution left in publishable content**
Line 23 — `This changelog draft was created by an automation.` — sits as plain body text between content paragraphs. If this file drives tweet copy directly (or is copy-pasted as-is), that sentence will appear publicly. It reads as internal metadata that should either be moved to an HTML/Markdown comment or removed before this ships.
### Issue 2 of 2
apps/marketing/content/changelog/2026-07-19-terminal-restore-failed-agents.mdx:6
**Unresolved TODO for hero and inline screenshots**
The JSX comment on line 6 calls out that hero and inline shots (terminal restore, red "failed" agent dot, ⌘F in Changes) haven't been captured yet. While `{/* … */}` comments don't render in MDX output, the weekly changelog ships entirely image-free while the companion performance post has six charts. Worth confirming whether the missing screenshots are intentional for this merge or a follow-up is planned.
Reviews (1): Last reviewed commit: "docs: enlarge before/after labels on per..." | Re-trigger Greptile
|
|
||
| Also this week: a terminal connection popover that tells you why a pane dropped, Clear Status now unsticks a wedged agent, adding an agent auto-creates its launcher everywhere, better auto-named workspaces when you don't run Claude or Codex, and refreshed Cursor models. | ||
|
|
||
| This changelog draft was created by an automation. |
There was a problem hiding this comment.
Automation attribution left in publishable content
Line 23 — This changelog draft was created by an automation. — sits as plain body text between content paragraphs. If this file drives tweet copy directly (or is copy-pasted as-is), that sentence will appear publicly. It reads as internal metadata that should either be moved to an HTML/Markdown comment or removed before this ships.
Prompt To Fix With AI
This is a comment left during a code review.
Path: apps/marketing/content/changelog/2026-07-19-tweet.md
Line: 23
Comment:
**Automation attribution left in publishable content**
Line 23 — `This changelog draft was created by an automation.` — sits as plain body text between content paragraphs. If this file drives tweet copy directly (or is copy-pasted as-is), that sentence will appear publicly. It reads as internal metadata that should either be moved to an HTML/Markdown comment or removed before this ships.
How can I resolve this? If you propose a fix, please make it concise.Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
| date: 2026-07-19 | ||
| --- | ||
|
|
||
| {/* TODO: hero + inline shots — capture on a machine that can launch the desktop app (background-terminal restore, red "failed" agent dot, ⌘F in the Changes pane) */} |
There was a problem hiding this comment.
Unresolved TODO for hero and inline screenshots
The JSX comment on line 6 calls out that hero and inline shots (terminal restore, red "failed" agent dot, ⌘F in Changes) haven't been captured yet. While {/* … */} comments don't render in MDX output, the weekly changelog ships entirely image-free while the companion performance post has six charts. Worth confirming whether the missing screenshots are intentional for this merge or a follow-up is planned.
Prompt To Fix With AI
This is a comment left during a code review.
Path: apps/marketing/content/changelog/2026-07-19-terminal-restore-failed-agents.mdx
Line: 6
Comment:
**Unresolved TODO for hero and inline screenshots**
The JSX comment on line 6 calls out that hero and inline shots (terminal restore, red "failed" agent dot, ⌘F in Changes) haven't been captured yet. While `{/* … */}` comments don't render in MDX output, the weekly changelog ships entirely image-free while the companion performance post has six charts. Worth confirming whether the missing screenshots are intentional for this merge or a follow-up is planned.
How can I resolve this? If you propose a fix, please make it concise.
🧹 Preview Cleanup CompleteThe following preview resources have been cleaned up:
Thank you for your contribution! 🎉 |
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@apps/marketing/content/changelog/2026-07-19-performance-memory-and-load.mdx`:
- Around line 12-22: Update the TL;DR headings and descriptions in the changelog
to state only outcomes supported by the listed measurements: memory reduction,
lower git-related stall duration, and reduced port-detection scanning. Remove or
qualify claims about fewer crashes, eliminating hitching, quieter fans, and
better battery life unless corresponding measurements are added.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 80e943f2-598f-403f-9131-9063097987a5
📒 Files selected for processing (2)
apps/marketing/content/changelog/2026-07-19-performance-memory-and-load.mdxapps/marketing/content/changelog/2026-07-19-performance-tweet.md
🚧 Files skipped from review as they are similar to previous changes (1)
- apps/marketing/content/changelog/2026-07-19-performance-tweet.md
| - **Less memory, fewer crashes.** Sessions with lots of terminals stay responsive and are far less likely to white-screen, even on 8 or 16 GB machines. | ||
|
|
||
| Renderer memory 1,127 → 838 MB (−26%), JS heap 581 → 286 MB (−51%), 4× fewer GPU contexts. | ||
|
|
||
| - **No more hitching from git.** Switching branches or changing lots of files at once no longer stutters the app. | ||
|
|
||
| Worst-case stall on the UI thread 82.3 → 28.0 ms (3× smaller), typical 24.5 → 9.5 ms. | ||
|
|
||
| - **Quieter in the background.** Port detection uses far less CPU, so quieter fans and better battery, even with many terminals open. | ||
|
|
||
| Background scans 15 → 1 per cycle (15× fewer), idle terminals scan every 30s instead of every 2.5s (12× less). |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win
Align the TL;DR claims with the measured results.
The provided metrics demonstrate reductions in memory, stall duration, scan count, and scan cadence, but they do not establish fewer crashes, eliminate hitching, or prove quieter fans and better battery life. Please qualify these statements or include the corresponding crash and power measurements.
Suggested wording
-- **Less memory, fewer crashes.** Sessions with lots of terminals stay responsive and are far less likely to white-screen, even on 8 or 16 GB machines.
+- **Less memory under heavy terminal loads.** Tested sessions used less renderer and JavaScript-heap memory.
-- **No more hitching from git.** Switching branches or changing lots of files at once no longer stutters the app.
+- **Less hitching from git.** In testing, worst-case UI-thread stalls fell from 82.3 ms to 28.0 ms.
-- **Quieter in the background.** Port detection uses far less CPU, so quieter fans and better battery, even with many terminals open.
+- **Lower background work from port detection.** Background scans fell from 15 to 1 per cycle, with slower polling for idle terminals.📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| - **Less memory, fewer crashes.** Sessions with lots of terminals stay responsive and are far less likely to white-screen, even on 8 or 16 GB machines. | |
| Renderer memory 1,127 → 838 MB (−26%), JS heap 581 → 286 MB (−51%), 4× fewer GPU contexts. | |
| - **No more hitching from git.** Switching branches or changing lots of files at once no longer stutters the app. | |
| Worst-case stall on the UI thread 82.3 → 28.0 ms (3× smaller), typical 24.5 → 9.5 ms. | |
| - **Quieter in the background.** Port detection uses far less CPU, so quieter fans and better battery, even with many terminals open. | |
| Background scans 15 → 1 per cycle (15× fewer), idle terminals scan every 30s instead of every 2.5s (12× less). | |
| - **Less memory under heavy terminal loads.** Tested sessions used less renderer and JavaScript-heap memory. | |
| Renderer memory 1,127 → 838 MB (−26%), JS heap 581 → 286 MB (−51%), 4× fewer GPU contexts. | |
| - **Less hitching from git.** In testing, worst-case UI-thread stalls fell from 82.3 ms to 28.0 ms. | |
| Worst-case stall on the UI thread 82.3 → 28.0 ms (3× smaller), typical 24.5 → 9.5 ms. | |
| - **Lower background work from port detection.** Background scans fell from 15 to 1 per cycle, with slower polling for idle terminals. | |
| Background scans 15 → 1 per cycle (15× fewer), idle terminals scan every 30s instead of every 2.5s (12× less). |
🧰 Tools
🪛 LanguageTool
[style] ~12-~12: Consider using a synonym to be more concise.
Context: ... memory, fewer crashes.** Sessions with lots of terminals stay responsive and are far l...
(A_LOT_OF)
[style] ~16-~16: Consider using a synonym to be more concise.
Context: ...m git.** Switching branches or changing lots of files at once no longer stutters the ap...
(A_LOT_OF)
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@apps/marketing/content/changelog/2026-07-19-performance-memory-and-load.mdx`
around lines 12 - 22, Update the TL;DR headings and descriptions in the
changelog to state only outcomes supported by the listed measurements: memory
reduction, lower git-related stall duration, and reduced port-detection
scanning. Remove or qualify claims about fewer crashes, eliminating hitching,
quieter fans, and better battery life unless corresponding measurements are
added.
What
The performance changelog for the week of 2026-07-19, plus its launch tweet and six before/after charts.
2026-07-19-performance-memory-and-load.mdx— a focused perf post. Every number is a real before/after measurement, not an estimate. Each section has a short write-up, a detail table, and two brand-styled charts:psscans 15 → 1 per cycle (15× fewer), idle cadence every 2.5s → 30s. Direct harness A/B against the pre-fix source.Charts follow the site's
DownloadButtonstyling (outlined brand fill, orange text, hard edges) on the dark surface tokens.Also in this PR
draftflag for changelog entries (draft: truein frontmatter → never renders on the site, list, RSS, or OG). Small filter inchangelog.ts.2026-07-19-terminal-restore-failed-agents.mdx) + its tweet, shipped as a draft so it stays in the repo but doesn't render. Flip the flag to publish it later.Notes
bun run lintpass. MDX compiles vianext-mdx-remote/serialize; all image references resolve.Summary by CodeRabbit