Joe Li

@Stanford | prev: @Cognition @NousResearch

[Joe Li @JoeLi5050]

@aaravwattal Wow this work is amazing aarav! Excited to see your future work in C2C interconnects

[Joe Li @JoeLi5050]

@AdamZweiger @thinkymachines @MIT strong

[Joe Li @JoeLi5050]

@cognition congrats!! 🔥🔥

[Cognition @cognition]

We’re releasing SWE-1.6, our best model in both intelligence & model UX. SWE-1.6 matches our Preview model on SWE-Bench Pro while dramatically improving on various behavioral axes.

It’s available today in Windsurf in two modes: free tier (200 tok/s) and fast tier (950 tok/s).

[Joe Li @JoeLi5050]

@RyanNeverWrong @weswinder @NousResearch 😂💯

[Joe Li @JoeLi5050]

@AdamZweiger Is this alpha????

[Aarav Wattal @aaravwattal]

🏆 We won Best Hardware Hack @hackwithtrees! (w/ @_apple_314 @slimtun3 @andgiuri)

We built Claude Code for hardware, and it shipped a 16,000-line bare-metal OS on Pi Zero W in 12 hours.

See more 👇

[Markus Zhang @photon_mz]

Over the weekend, I managed to beat the NanoGPT world record by 0.3s, by untying Value Embeddings (VE) 😊

There’s a growing trend in sparse embeddings (Deepseek engram, Meta STEM), and @karpathy himself has extensively ablated VEs in nanochat

[Joe Li @JoeLi5050]

@photon_mz @karpathy very cool!

[Joe Li @JoeLi5050]

@AadvikVashist appreciate you aadvik 🙏

[Weights & Biases @wandb]

This is what incredible ML research looks like.

Congrats to @Teknium, @JoeLi5050, @rogershijin & the rest of the cracked out Nous Research team on pushing open-source code reasoning forward.

[Joe Li @JoeLi5050]

@NousResearch 🔥🔥🔥

[Joe Li @JoeLi5050]

Enjoyed working on this a lot! Check out the blog as well.

https://t.co/zjmhSnBRup

[Nous Research @NousResearch]

Introducing NousCoder-14b, a competitive olympiad programming model.

Our latest blog details the full findings from extensive experiments and logs with the full stack released - the RL environment, benchmark, and harness built in Atropos, all fully reproducible with our open training stack.

NousCoder-14b was post-trained on Qwen3-14B by researcher in residence @JoeLi5050 using 48 B200s over the course of 4 days, our Atropos framework, and @modal's autoscaler. It achieves a Pass@1 accuracy of 67.87%,+7.08% over Qwen's baseline accuracy using verifiable execution rewards.

[Joe Li @JoeLi5050]

@radi_cho Wow!

[Joe Li @JoeLi5050]

@radi_cho @Google @NanoBanana @googlemaps so cool

[Bleu AI @BleuAI_]

Introducing Bleu AI.

Have an AI app idea but no way to build it? Build it in an instant with Bleu AI's visual workflow builder.

Access to Grok, GPT, Claude, and dozens of AI models.

Reply "buildbleu" for 1M Grok tokens.

Visit https://t.co/9hdsavuOci today.

[Joe Li @JoeLi5050]

@BleuAI_ buildbleu

[Joe Li @JoeLi5050]

@SallyHZhu Great work!

[Joe Li @JoeLi5050]

@RyanNeverWrong Very cool

[Joe Li @JoeLi5050]

@BleuAI_ @grok explain what this is?