HNHacker News
TopNewBestAskShowJobs

lawrencechen

1,388 karma · joined March 30, 2021

https://cmux.com

https://x.com/lawrencecchen

lawrence@cmux.com

submissionscomments
lawrencechen··on Show HN: cmux - Ghostty-based terminal with vertical tabs and notifications
Mobile interface is definitely nice. Planning on adding iOS app since libghostty works there too! And I imagine that having your main terminal app be synced directly to your phone must be nice, though it doesn't solve the problem of closing my laptop.

Would love to hear what other features have been particularly beneficial to your dev style too. Some directions I'm interested in is having everything be programmable; so my coding agent can set up workspaces for me, click through browsers to test things, etc. And having a main Claude Code manage subagents that have their own easily visible terminal windows.

lawrencechen··on Show HN: cmux - Ghostty-based terminal with vertical tabs and notifications
Thank! I personally have caps lock mapped ctrl... but open to suggestions! Since it's hard to handle both horizontal and vertical tabs.
lawrencechen··on Show HN: cmux - Ghostty-based terminal with vertical tabs and notifications
Thanks! Will add drop target and sidebar coloring.
lawrencechen··on Show HN: cmux - Ghostty-based terminal with vertical tabs and notifications
Upstreaming into Ghostty would be very difficult as it's not actually a fork, I just used libghostty under the hood.
lawrencechen··on Show HN: Skill that lets Claude Code/Codex spin up VMs and GPUs
I also personally prefer running agents locally instead of in the cloud. For some reason, it feels easier to steer Claude Code when it's running in my terminal vs steering something in the cloud. Maybe part of it is the latency from typing into ssh'd TUIs, and this is something that a GUI can solve... but I still feel more at home with Claude Code/codex in the CLI vs something like Claude Code Web/Codex Cloud. Part of it is likely reliability and "time to first token that AI responded that I can see." But local has tradeoff of conflicting ports/other resources, lag (maybe it's time to upgrade my M1 Max 64gb...), and slight latency incurred since LLM calls have slightly more network latency.

Curious to hear more about your local orchestration platform, how did you solve resource sharing (mainly ports for web stuff tbh)? Or is it more intra-task vs inter-task parallelism?

lawrencechen··on Show HN: Skill that lets Claude Code/Codex spin up VMs and GPUs
Railway is awesome! Pretty different use cases though - Railway's MCP is for deploying and managing persistent services (git-push-to-deploy). CloudRouter is about ephemeral sandboxes: the agent spins up a throwaway VM, does its work, and tears it down.

We are definitely inspired by Railway though!

lawrencechen··on Show HN: Skill that lets Claude Code/Codex spin up VMs and GPUs
Fair enough, is this mainly for running services or do you use it for dev loop too?
lawrencechen··on Show HN: Skill that lets Claude Code/Codex spin up VMs and GPUs
Right now, all usage is routed through us, hence `cloudrouter login` is required for now. Plan on adding bring-your-own cloud/key, but having something that's fast to setup is hard since we need to pre-build a template that includes VNC, browser, VSCode, worker daemon, etc...
lawrencechen··on Anthropic's original take home assignment open sourced
codex cli + gpt-5-2-codex-xhigh got to 1606 with the prompt "beat 1487 cycles. go." ~53 minutes.
lawrencechen··on Creators of Tailwind laid off 75% of their engineering team
They were perfectly positioned to build a Lovable/Bolt/Replit back in the day... might not be too late now either.

They could sell training data too. Though, UIs are relatively solved. But great UIs and criticizing UIs aren't.

Learned a lot from Refactoring UI, and I know (from trying) that it's impossible to make a code review bot based on out of the box sota models today. Vision capabilities are lacking here, and I can see demand for more data here. And Adam's taste likely fits well here.

lawrencechen··on Show HN: Cancer diagnosis makes for an interesting RL environment for LLMs
I wonder if navigation plays a significant role in performance. If you just randomly select 15 frames (presumably with interesting pixels), will the model perform similarly well?
lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
cmux desktop app currently requires signing in to GitHub. We will build out better support for local repositories and remove sign in requirement soon.
lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
If your goal is to make something useful, I think the fastest way is probably to build a CLI only version since you can theoretically render heatmaps and make a task manager in a CLI form factor. And your background in Java/C# helps here.

Use Claude Code or Codex for everything, learn how to prompt well. >90% of cmux and 0github.com was written by LLMs. Most of it was just me asking the LLM to implement something, testing it to see if it works, and if it doesn't, I'll ask the LLM to write logs, and I'll paste the logs back to the LLM. Ask gpt-5-pro for architecture choices, like what tech/dependencies to use.

But if your goal is to learn React, I'd recommend going through the official getting started documentation, it's pretty good.

lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
Yeah, adding a context gathering step is a good idea. Our original approach used codex cli in a VM, so context gathering was pretty comprehensive. We switched to a more naive approach due to latency, but having a step using a smaller model (like SWE-grep) could be a nice tradeoff.
lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
Public repos shouldn't require being signed in.

Just tested these example links in incognito and seemed to work?

https://0github.com/manaflow-ai/cmux/pull/666

https://0github.com/stack-auth/stack-auth/pull/988

https://0github.com/tinygrad/tinygrad/pull/12995

https://0github.com/simonw/datasette/pull/2548

> you've disabled logging issues on the repo

Sorry, wasn't aware. Turning it on right now. EDIT: https://github.com/manaflow-ai/cmux/issues seems to be fine?

lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
Yeah, heatmapping the diff before creating a PR would need tighter IDE integration. We're working on cmux for this purpose. It's kinda an IDE, and it lives in the same repo: https://github.com/manaflow-ai/cmux.

After we add the heatmap diff viewer into cmux, I expect that I'll be spending most of my time in between the heatmap diff and a browser preview: https://github.com/manaflow-ai/cmux/raw/main/docs/assets/cmu...

lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
Good suggestions! Will make it more URL friendly.

> Adding a comment to explain why the downgrade is done would have resulted in not raising the issue?

Trying it out here with a new PR on same branch: https://0github.com/manaflow-ai/cmux/pull/809

Will check back on it later!

EDIT: seems like my comment online 62 got highlighted. Maybe we should surface the ability edit the prompt.

lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
Happy to hear!

> Is this using some sort of measure of distance between the expected token in this position vs the actual token?

The main implementation is in this file: https://github.com/manaflow-ai/cmux/blob/main/apps/www/lib/s...

EDIT: yeah it's just a LLM prompt haha

Just a simple prompt right now, but I think we could try an approach where we directly see which tokens might be hallucinated. Gonna try to find the paper for this idea. Might be kinda analogous to the "distance between the expected token in this position vs the actual token."

lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
Yeah this is honestly pretty expensive to run today.

> I’m not sure an LLM can really capture project-specific context yet from a single PR diff.

We had an even more expensive approach that cloned the repo into a VM and prompted codex to explore the codebase and run code before returning the heatmap data structure. Decided against it for now due to latency and cost, but I think we'll revisit it to help the LLM get project context.

Distillation should help a bit with cost, but I haven't experimented enough to have a definitive answer. Excited to play around with it though!

> which parts of the code change most often or correlate with past bugs

I can think of a way to do the correlation that would require LLMs. Maybe I'm missing a simpler approach? But agree that conditioning on past bugs would be great

lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
Getting rate limited by GitHub, gonna add caching here as well. Temporary workaround is to sign in manually and return to example page: https://0github.com/handler/sign-in
lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
Yeah we definitely want to make the gradient and colors configurable.

What form factor would make the most sense for you? Maybe a a cli command that renders the diff in cli or html?

lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
Shoot, we should have caching in place already. Taking a look now
lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
https://github.com/manaflow-ai/cmux/blob/main/apps/www/lib/s...

This file has most of the logic, the commit you linked to has a bunch of other experiments.

> look at your prior reviews and try to learn your style.

We're really interested in this direction too of maybe setting up a DSPy system to automatically fit reviews to your preferences

lawrencechen··on Show HN: I made a heatmap diff viewer for code reviews
Currently tooltips are shown when hovering on highlighted words. Need to make it visible on mobile though. Was wondering if you were thinking of another way to show the labels besides hovering?
lawrencechen··on Why warm countries are poorer
Related question I've been asking my friends: if you were forced to get rid of air conditioning or democracy, which would you choose?
lawrencechen··on The Bitter Lesson Is Misunderstood
In the bitter lesson essay [0], the word "data" is not mentioned a single time.

The author fundamentally misunderstands the bitter lesson.

[0] https://www.cs.utexas.edu/~eunsol/courses/data/bitter_lesson...

lawrencechen··on Launch HN: Issen (YC F24) – Personal AI language tutor
Congrats on the launch!

I tried the Japanese track. I'm a total beginner and the first lesson wasn't helpful at all. The AI asked about maybe mixing up Japanese<>English, but it didn't actually follow through. It either spoke fully in Japanese or fully in English. Maybe this is a standard practice for language lessons? I remember going to the first day of French class in a community college, and the teacher only spoke French, which was extremely overwhelming. Perhaps it's the standard way of teaching? Even if it is, I'm not sure if it works when compressed down to the shorter times I see myself opening the app.

lawrencechen··on Snorting the AGI with Claude Code
I think he means it's not economically sound to use it via API
lawrencechen··on Show HN: Puck 0.18 adds drag-and-drop for CSS grid and Flexbox
Great work!

Is it possible to resize grid rows/grid cols directly in the editor? Similar to the bottom-right handle in react-grid-layout?

https://react-grid-layout.github.io/react-grid-layout/exampl...

lawrencechen··on Launch HN: Retell AI (YC W24) – Conversational Speech API for Your LLM
thanks!
← PreviousPage 2 of 3Next →