Graham Neubig

Our paper is now ungoogleable 😭 https://arxiv.org/abs/2507.06134 (but congrats to NVIDIA on the effort)

I sadly won't be participating in #COLM2026 (instead will be teaching the CMU AI Agents course), but my wonderful students will be there. Say hi to them, especially the ones on the job market this year.

just a reminder that if you're complaining about {openai/claude} prices, deepseek-v4.1-flash and glm-5.3-flash can do 90% of your work for you. there's a place and time for frontier intelligence, but you certainly don't need to pay a boatload for it.

Hard to believe that one night I hacked together a silly non-functional JS app, and a couple years later we have 90k stars on github and over 500 contributors 😄

Impressive first release by Reflection! Good things come to those who wait, apparently 😀

Wow, very cool and timely benchmark for personal agents!

This is so true! At OpenHands we have have reached the "bonus stage" from the original post. Our process has been rebuilt around the assumption that if an issue is opened on the repo, we are guaranteed get a vibe-coded PR opened, so our defense is: (1) make sure that all the issues are maximally clear through automated issue triage, and (2) automate code review so that we only accept high-quality contributions. All of these are powered by OpenHands automations: https://github.com/OpenHands/extensions

New quick explainer video on how to set up automated tasks using the OpenHands agent canvas: https://www.youtube.com/watch?v=3FzRrT3HH3w We've also used this to set up a "software factory" for the OpenHands project as well, hope to post a more comprehensive video about that soon.

I bought a B300 machine for my lab at CMU. I was thinking of using it to do research on LLMs, but now I'm thinking of whether I should just use it to serve deepseek-v4.1-flash tokens with the new deepseek inference code and fund the entire rest of my lab that way 😅

Don't sleep on single agent systems https://www.openhands.dev/blog/dont-sleep-on-single-agent-systems

Tried out Muse Wonder on a tricky task (figuring out which parts of the OpenHands Agent Canvas are hard for users to use) and it did a great job, impressed! And as someone who's used to gpt models, the conciseness of the verbiage was pretty nice too 😀 Great job by @AIatMeta !

Amazing. For those who have used DeepSeek before, you know that the inference that they have on the official provider is absurdly good+cheap. This is a big release.