Here is the full talk:

Is there a YouTube link instead please? It's a much better video player.

@Scobleizer This is the part people underestimate. AI raises the floor fast, but it can also flatten the ceiling if everyone starts outsourcing the same parts of thinking.

Is there a short form version of it with subway surfers gameplay below?

the moment you said "alien tool", in the first 30 seconds of the video convinced me already that you have nothing to offer

Also on Youtube, looks like a good watch. BRB.

I think you did a good job in this presentation one thing I think you undersold is the point about generic thoughts from the model at the end or that its getting worse at writing - thats a feature of the context provided and the depth of trajectory at which the question was sampled, not just an innate problem with LLMs. It's unsolvable in the way you mean - but the solution is to just build more interesting divergent trajectories if you want diversity when you're sampling.

Thanks for sharing

@grok summarize this talk for me, main findings as bullet points (this is a joke)

Can't wait to check it out Thanks for sharing the link

I'm glad this popped in my feed! I was worried I would miss the link. Thanks for sharing.

@bot summarize the content for me

@BrianRoemmele

Thanks! Slides available?

So i have tried socratic but claude just gives the answer way too easy in the question. Do you have any prompts that you use. Or i mean something batle tested?

Is it on YouTube?