The last IMO problem AI could not solve. For when you want to sit down for a longer problem-solving saga, here's the full video:

Wouldn’t it just be a line from corner to corner

Yeah, but AIs don't solve the problems now because they are so smart, but because the answers are now part of their training data. I could also state I can solve all of them, but that doesn't mean I am smart if I read the solutions beforehand

Great video, I just had one question. Any connection to sudoku?

I normally never comment but you saying that there is a longer more thorough explanation available is great keep up the good work i appreciate your time and effort :)

I like how he makes long video interesting to watch

Of course I will. Grant, since your series on Neural Nets, a lot has happened on the LLM front. So much advancement in the core architecture of advanced LLMs. Are you looking to dive deeper into today's state of the art ? I wish you did!

AI solved five of six IMO problems, but Problem 6 exposed the remaining gap: sometimes the hardest step is finding the right idea, not carrying out the proof.

This has such cool animations, I think its an upgrade from other videos

The hard minimum for an NxN grid is N tile placements, since if you imagine a free space above and below an X, a rectangle can't hit both of these squares without including the X, therefore, you cannot place a tile such that it fills up more than 1 row at a time. N rows.

I learn more from the miss than from the gold medal thread

When the two terms starting with "L" showed up to prove sqrt(N), I felt like someone revealed a magic trick on Penn and Teller. Insane to think this problem boils down to that (/.\) Great video :D

If the quality, complexity, and polish of animation continue to improve at this rate, I'll soon experience Stendhal syndrome.

This is a beautiful journey.

the stronger imo benchmark uses hidden variants fixed prompts and one attempt

@grok solve this.