Also I think there are multiple distinct approaches to build an AGI within this

Then we 1) Want to compress the knowledge effectively in a near lossless way 2) Cmpose the pieces of knowledge effectively to approximate the wanted solution (chollet thinks of this as program search)
Most (all?) AI systems can be abstracted in to these 2 things

Thanks, I missed that

In your statement you mention analysing the results that o3 wasn't able to solve. Where is the best place to post findings on this?