Just completed a full end to end run of the full 113 task DeepSWE benchmark on ox-alpha. The Rumored ~80% pass rate is completely incorrect. Actual benchmark result is 58.4%, landing the model at almost identical performance to Claude Opus 4.8 (59%) 🧵
https://t.co/8VnZYTb1xv
We made a harness that allows agents to disassemble, play, and inspect running memory state of games at #grokathon
Using the harness, Grok Build reverse engineered GameBoy Breakout from the raw binary. Give it a shot here:
https://t.co/QisxGCxz9b
Germany is the world’s third largest economy after the US and China, and the economic powerhouse of Europe. We are thrilled to have Canada and Germany as deep strategic partners at our back! 🇨🇦 🇩🇪
i'm so excited to finally share what we've been working on @rhodaai! taking lessons from success in LLMs, we identify autoregressive video generation as a scalable objective for training robot policies.
1/n