Hacker News Reader

It does not say anything about real use cases. It performs better and "reason" better than o1-preview and o1. But I was expecting some real-life scenarios when it would be useful in a way no other model can do now.

View on HN

I imagine the system prompt is something along the lines of, 'think about 10% harder than standard O-1'

for every tier that costs 10x more than the previous, they add a "very" to the "You are a very, very, very smart AI"

More like iterations and depth of tree of thoughts search 3× in the pro mode.

The point of this tech is that with scale it usually gets better at all of the tasks.

Reply on news.ycombinator.com