This is a free preview of a premium episode. To subscribe to Down Round Premium and get access, head here.

After 2 months of cringe-posting references to strawberries, OpenAI have released their latest model: o1. And AI heads are losing their minds.

Ostensibly, it's intended for the kind of thorny math and reasoning problems that LLMs tend to struggle with. The guts aren't public, but it seems to use chain-of-reasoning prompts to really force itself to 'think'. 

So, is it any good? Is this a huge step change on the path to AGI. We investigate.

See omnystudio.com/listener for privacy information.