Discussion about this post

User's avatar
Dominique Paul's avatar

Interesting! One thing I'd say is missing are low-cost post-training methods:

Replicating DeepSeek R1 in the CountDown game for $30:

https://x.com/jiayi_pirate/status/1882839370505621655

Beating O1 at math for $4,500:

https://x.com/Yuchenj_UW/status/1889387582066401461

Expand full comment

No posts