Sahil Verma

@hazhubble @butcher_jasper

OFFER: if you want to come to SF but budget is stopping you, i want to offer you FREE housing in our hacker house! we have the best views in SF, fancy chairs and monitors, 1gbps internet, and queen size beds waiting for you if you know someone who should be in SF tag them below DMs open

@Yuchenj_UW where do you find about these meetups?

Beyond raw model capability, the real gap in coding tools is the harness. Now that 500k+ lines of Claude Code are out there, every model lab and AI coding startup, including open-source AI labs, will study it and close that gap fast. SF already has Claude Code source walkthrough meetups lol.

@ArmenAgha 100%, thank you

The Codex app is really nice, but I really wish it supported Remote-SSH. Most of my work requires GPUs on remote clusters/envs... adding this would make me move to Codex full-time.

@reach_vb @sama Unfortunately it doesn't work with remote ssh server code, would that change soon?

@sama YESSSS! and people some to be loving their 2x rate limits too! Team is already hard at work to ship an update too Download the App here: openai.com/codex

More than 200k people downloaded the Codex app in the first day. And they seem to love it. CODEX FTW!

@Tim_Dettmers This is so good, Tim. Very interesting that you read about Krishna and Arjuna!

My newest blog post is about my experience of using coding agents in the past 8 months to automate my own work as much as possible. This post introduces concepts I found useful and has stories of successes and failures when I used agents in real settings. timdettmers.com/2026/01/13/use…

@paraschopra @lossfunk @butcher_jasper

We're making a 2026 calendar for talks at @lossfunk. Who is operating at the frontier of AI and the changing world? Artists, thinkers, doers, musicians, builders - please tag anyone who you think deserves to be seen, and we will reach out.

🚀 Nemotron 3 Nano is live! Had a blast post-training this model with a cracked team. Its strong for its size, and highly efficient at inference. And true to @nvidia's open release style: weights (BF16/FP8/base) + training recipes + code + datasets. HF: huggingface.co/collections/nv… Blog + Nano tech report: nvda.ws/48RusVt

@liweijianglw You are leaving a real heritage @liweijianglw!

Taking a full week off to recover from NeurIPS heritage (aka COVID), and just want to say it was an absolutely wonderful conference experience. One of the biggest highlights for me was talking with junior students. I still remember how eye-opening my first Grace Hopper Conference was years ago. It completely changed how I saw computer science, especially coming from a liberal arts college without research resources. That’s why I really really want to spend time talking with junior folks, in case these moments help ignite their passion too, like once it happened to me. Two moments especially stayed with me: (1) A junior student came up to me to thank me for helping revise their PhD application a few years ago, not only did they get into great programs, they’re now helping younger students apply as well. (2) Another junior student shared that my work on Delphi inspired her to start NLP research two years ago, and that reading it during a cold winter really lit up her world. Now they're applying to PhD programs, and have loads of passions to demystify human minds with AI. Hearing this meant so much to me. If my work and time can have even a small positive impact, it feels like a way of paying forward the mentorship I’ve been fortunate to receive. And I'll forever carry this on in my career!

@r_krishna3 @naturemethods Congratulations @r_krishna3 and @butcher_jasper

Today, we report a method for design of active enzymes, RFdiffusion2, in @naturemethods. For the first time, we are able to design enzymes with native-range catalytic activity. We also are releasing our next frontier model, RFdiffusion3, code 👇

Sarah is great, go work with her!

[1/9] While pretraining data might be hitting a wall, novel methods for modeling it are just getting started! We introduce future summary prediction (FSP), where the model predicts future sequence embeddings to reduce teacher forcing & shortcut learning. 📌Predict a learned embedding of the future sequence, not the tokens themselves

Hands down Hila is one of the best advisors out there, you would be lucky to work with her! (as was I :) )

Very excited to share our paper "De novo Design of All-atom Biomolecular Interactions with RFdiffusion3", now on BioRXiv. biorxiv.org/content/10.110… 1/n

RFDiffusion3 generates all atom bound conformation, making it significant for flexible targets like DNA. An excellent teamwork to achieve something impossible by any one of us in just few months. @butcher_jasper @r_krishna3 biorxiv.org/content/10.110…

@jbhuang0604 your videos are really good!

YouTube hides dislike counts from the public, but creators can still see them. It’s tough to face, as each video takes many hours of thought and work. BUT, it's also a reminder to just keep improving!

@fr0gger_ is the first screenshot real?

Prompt Injection is one of the first attack vectors used to exploit weaknesses or bypass behavior in AI models. Here is an illustrated thread with 5 different prompt injection techniques 👇

Glad to share that our paper was accepted the main EMNLP 2025 Conference! x.com/Sahil1V/status…

@gowthami_s not too late to change

There is liberation that comes with belief. Sometimes I wish I’m religious.

@yanaiela Congratulations Yanai!!

I’m excited to share that I'm joining Bar-Ilan University as an assistant professor!