🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

This is the improvement Claude Code needed to be great. Just keep going! I don’t care how you do it or what the context looks like. Looking forward to trying it.

Hmm…. Should I do a swebench run with this?

Why am I debugging AI’s code instead AI debugging my code?

Does discouraging external linking on X make it a better model training corpus?

Legend

CoreWeave is the first cloud provider to submit MLPerf Inference v5.0 results for @NVIDIA GB200 GPUs, achieving a 4X per-chip performance improvement over H200 GPUs. We are committed to delivering the fastest and most efficient AI infrastructure. hubs.la/Q03fBklc0

OpenAI’s o4 just showed that multi-turn tool use is a huge deal for AI agents. Today, we show how to do the same with your own agents, using RL and open-source models. We used GRPO on only 100 high quality questions from the BFCL benchmark, and post-trained a 7B Qwen model to

I got my first startup office space in SF from Aneel for free back in 2007. Highly recommended. He's an awesome host!

New Evals API I’m excited to share a new API for logging evals with W&B Weave. EvaluationLogger - log_prediction - log_score - log_summary Our design goal for this API was to get out of your way and build the most flexible eval API out there, inspired by wandb.log, which our

🚀 Meet ART·E—our open-source RL-trained email research agent that searches your inbox and answers questions more accurately, faster, and cheaper than o3. Let's go deeper on how we built it. 🧵

We only need ONE example for RLVR on LLMs to achieve significant improvement on math tasks! 📍RLVR with one training example can boost: - Qwen2.5-Math-1.5B: 36.0% → 73.6% - Qwen2.5-Math-7B: 51.0% → 79.2% on MATH500. 📄 Paper: arxiv.org/abs/2504.20571

We’ve officially completed our acquisition of Weights & Biases , and I couldn’t be more excited. Combining CoreWeave high-performance AI cloud with W&B’s incredible developer tools unlocks new levels of innovation for our customers. Together, we’re building the next-gen AI cloud

BIG NEWS: Fastino raises $17.5M Seed to launch TLMs – Task-Specific Language Models that beat GPT on accuracy and latency. Led by jon chu at Khosla Ventures + joined by George K. Mathew at Insight Partners, agracias at @valorep, Scott Johnston (ex-Docker CEO), and Lukas Biewald (CEO of

Come hang out at our conference, Fully Connected, June 18-19 in SF!

Optimize for self-confidence, no external rewards needed. Beautiful work. I have a feeling this leads to models claiming to be self-aware.

the #1 use case for AI coding should be to vibe code interfaces to Look At Your Data

Excited to talk about ART and RL for agent reliability at the Weights & Biases Fully Connected keynote alongside Shawn Lewis. Thanks for the invite!

Wow, amazing results with ART!