rohitpaulk

We use RubyLLM in production too, the most elegant library in this space I've seen so far.

I also liked how they run the issue tracker. If you select "Feature Request", it makes you explain how you explored workarounds, why you believe it belongs in RubyLLM etc to prevent scope creep.

Same here. Loved the LTS thinking and Ember data being built in but hiring was a pain and ecosystem packages were often out of date.

Prior art: https://blog.can.ac/2026/02/12/the-harness-problem/

The whole idea that my RUNTIME contains code that a single human hasn't looked at does make me uncomfortable, but if this actually works without a ton of issues it's pretty remarkable.

Aside from how impressive the model is, the demos here are very well done! Quirky and short, unlike what we're used to from Anthropic and OpenAI.

Only a portion of these are "scandals", the rest are just usual startup failures.

That's 2 major layoffs this week (Coinbase being the other). Is there an underlying common reason for this? And is it indeed AI-driven productivity as both companies claim?

Nice one. Pleasantly surprised that this has nothing to do with Cursor the IDE

Been testing these via their "pool" agent. It's fast, and the agent adheres to the ACP spec pretty well (better than codex, opencode etc.) so it's a good experience in Zed.

Most of the comments here are clearly from people who haven't used GitButler. Try it out and it's a very sticky product, clearly superior workflow to vanilla Git.

Saw this on Twitter a few weeks ago, interesting approach.

The post links to the GitHub repo, but imo the website does a better job of explaining what it does: https://ataraxy-labs.github.io/weave/

Two questions I'd have expected the article to answer:

(a) how did they identify the employee, and (b) how come they weren't sent to jail

Surprised that none of the comments here are comparing this to Bootstrap.

This sounds like a good optimization task to give a long-running agent. Ask it to come up with experiments and maximize the % of successful edits.

In case anyone finds it useful, we (CodeCrafters) built a coding challenge as a companion to this book. The official repository for the book made this very easy to do since it has tests for each individual chapter.

Link: https://app.codecrafters.io/courses/interpreter/overview

Wild how this beats 2.5 Pro in every single benchmark. Don't think this was true for Haiku 4.5 vs Sonnet 3.5.

For those who haven't seen the actual issue that links to, I'd take a closer look. Pretty insane: https://github.com/actions/runner/issues/3792#issuecomment-3...

Always felt dependency updates are a perfect fit for AI agents:

(a) they’re broadly similar across companies,

(b) they aren’t time-sensitive, so the agent can take hours without anyone noticing, and

(c) customers are already accustomed to using bots here, just bad ones

I noticed this with Harvey / Legal.

Where do I buy one?

- Contribute to OSS - Do CodeCrafters

Great content! Sidenote for the Fly team: on mobile, the “sidenote” cards appear in the wrong order - they appear before the content instead of after.

Not exactly what you described, but https://labs.iximiuz.com/ is similar

Still involves friction. A more "seamless" way for Apple to do this would've been to license GPT-4's weights from OpenAI and run it on Apple Intelligence servers.

Definitely curious about how this works - signed up!

This is a great idea. We’ve used Storybook & Chromatic in the past to cook up something similar for our front-end apps. There was quite a bit of setup work involved though.

How does integrating this into a web app work? Do you hook into an existing test suite? Record sessions on staging?

Tried this out on a Ruby codebase and it generated Python tests: https://github.com/Codium-ai/cover-agent/issues/17. Is there any data available on whether this actually works?

Looks like GPT-4 is 3x slower on Azure compared to OpenAI?

(disclaimer: I work at CodeCrafters)

CodeCrafters is a bit light on Elixir support at the moment (only 2/8 challenges support it so far). We're planning on expanding Elixir support soon.

Currently, we're a great fit for Rust / Python / Go - we support a variety of challenges for those, and there's lots of example code from other users to review and learn from.

Wonder what the non-compete looks like for a role like this. Will Sam Altman be allowed to create an OpenAI competitor right away?