GROK 4.5 IS NOW ON THE FRONTIERCODE LEADERBOARD
Cognition just launched the FrontierCode leaderboard — a new benchmark specifically designed to track which AI models are writing code you’d actually merge into production.
Unlike many synthetic benchmarks, FrontierCode focuses on real-world usability, with full methodology and sample tasks publicly available.
Grok 4.5 is included in the rankings alongside other top models.
This is another step toward measuring what actually matters for developers: code that works, integrates cleanly, and solves real problems.
instead of watching 2 hours of Netflix tonight, watch this Stanford lecture
it's the clearest explanation I've seen of how LLMs like ChatGPT and Claude actually work
useful whether you've never touched AI in your life or have been using it every day for the past year
I took the key ideas and turned them into a practical guide on how to actually build LLM from scratch
find it below
Technology never ceases to amaze me. It's incredible how it connects us, simplifies our lives, and sparks creativity in ways we never imagined. Can’t wait to see what’s next!
This journey has been everyone’s journey. Thank you to all involved in the success of our mission, and for giving me the opportunity to carry everyone’s dreams into space. I’m filled with gratitude to be back on the planet!
New paper online in @naturemethods!
An optimized chemical-genetic method for cell-specific metabolic labeling of RNA
https://t.co/Z3JeXwYSCE
Thank you to reviewers and editorial staff for their very helpful comments and guidance!