Geometric Reasoning

See https://github.com/guilt/TinyToT which already builds a powerful reasoning model where you drop your traces as Markdown, it picks it up and runs. Ditto for data, so you have both separately and can refresh one and not mess up the other. Zero training, Zero RL. Code is there, try it yourself.

Someone's doing a company out of this? Great, let more people discover the truth.

Premise goes: our brains are infinitely more efficient at reasoning than current computational architectures. Presumably because our synapse permits allow for non linear connections in a dense space (grey matter). At a biotech I worked at the phrase that described this difference was “volumetric computation” vs “linear computation”; I can’t find a reasonable description of this online so I’m going to assume it was a house term.

This site seems to suggest the internal geometry of the connective/reasoning layer is more important that having more data. This causes me to think “graph” vs “array”.

Or yeah, maybe it’s just slop and some number of people will be creative and buy the hype ^_^

Anyway the website doesn't look very sloppy to me.