AI Friends

yudhister: OSWorld (at least, in its current form) doesn't quite fit: https://t.co/q72HTYKACY

OSWorld (at least, in its current form) doesn't quite fit: https:// x.com/G413N/status/2 026017035612996056?s=20 … right now we're pretty excited about game evals, like https:// arxiv.org/abs/2602.17594 or ARC-AGI 3. LLMs are pretty bad at these and we think we can do much better
x.com x.com x.com x.com t.co
DesignResearchSocial Threadx.comLearning