We're making Git hosting more reliable, performant, and scalable.
This post traces 20 years of Git infrastructure and explains how that history led us to design and operate our Git storage, Origin, as if it were a database.
https://t.co/UW7jHuItSX
@realchrisebert@lydiahallie Seeing the same with Opus 5. I’ve had better results treating verbose comments as a review problem (trim down) rather than trying to steer the implementation agent’s context hard enough to prevent them
If you've noticed comment bloati, it's measurable, and I don't think it's just chattiness. Posted these Python scripts to run both analyses on your own repo & CC session logs
https://t.co/MTFp2MfxYp
Especially curious whether the depth increase replicates!
Have noticed a significant change in Claude's comment density / quality as of late. Ran analysis on around 500 sessions & git history:
- 18% of all output from Opus 5 is now comments (>2x from 2 months ago on 4.8)
- 5x more 8+ line comment blocks at 200+ tool calls
Same shape in two places:
The longer a file has been worked, the denser it gets. 47% of my 8+ line blocks were built across multiple commits.
The longer a session runs, the denser its output. 0.82 → 4.20 blocks per 1k lines. No idea yet if that's one mechanism or two.
When I worked at Apple, I remember seeing code that was 15 years old. Every file had a comment at the top with the date and the author’s name. You’d open a file and read through it, sometimes touching a feature no one had touched in a decade.
Everything felt very permanent, like a house made of bricks. Everyone who made a brick put their name on it before slotting it into the house. Even in a massive company, you were proud of your bricks. Each one fit so satisfyingly into the bricks around it.
When someone was renovating a room you had worked on years ago, they’d reach out and say, “I found this brick with your name on it. Can you tell me what it was doing here?”
Sometimes old bricks would get taken out during a renovation, but they were usually kept in a pile in the corner to reuse in the future.
Now you can will whole houses and neighborhoods into existence. You don’t even see the bricks go in. They just tumble in from the sky. You walk into the new house and wave your hands to push walls back, throw up a chimney, or drop in a kitchen.
It's so fucking cool. But you forget what a brick feels like.
And it’s hard to know how sad to be about that. The houses are better, they go up faster, and building them increasingly feels like magic.
Still, every once in a while you miss holding a brick.
@chandrxn Find something that exists that inspires you and chip in. If nothing inspires you, start something that would. Find others like this. Repeat until career is done.
Introducing uploads
The missing upload command for coding agents
→ One command: `uploads put`
→ Screenshots are hosted, nothing is committed to your repo
→ Capture while you work, branch detection — no PR needed yet
→ PR opens with one comment for all visuals, labeled
→ Before/after pairs rendered side by side
→ Images update in place on revision
→ CLI, MCP + GitHub App
→ Works with any agent
→ Open source (cloud or self-host)
@1jehuang thanks for jcode - the performance is incredible!
Wondering if you are planning full plug-and-play with existing CC setup? It looks like skills/mcp already carry over, but some config/context (see screenshot) is silently ignored. Could we get a warning in the interim?
@lawrencecchen great to see that your token usage is so high, but the software you’re releasing has so many bugs that its crashing developers’ machines on a regular basis. Many cmux releases are followed 24-48 hours with hotfix patches that resolve a dozen critical issues.
Would you consider slowing down on shipping new features in favor of building more stable, performant terminal? It would be great to use this tool without its memory footprint of, at times, 20gb+ for just a handful of open workspaces.
@imbabybrooklyn@Teknium For as much as I have tried to make this work, I always return to a stateless workflow. Even with good orchestration e.g. N8N, there are too many edge cases (that can also result in expensive loops and token leaks, no thank you)
I've started a new company: @superlogical! We're going to begin by building a terminal multiplexer. The entire vision is much larger, but the multiplexer is the foundation. Sign up for the newsletter to get beta access and devlogs (product updates only I promise).
https://t.co/XJuqufkmCd