o3 pro is a BEAST... one-shots Apple's "Illusion of Thinking" test

To be fair "Apple Illusion" would certainly have been a better name than "Apple Intelligence" 😂

o3-pro might be writing code, executing it, evaluating it, fixing bugs, and then repeating the cycle until it reaches some level of confidence. That might be why it was able to one-shot Hanoi, because it knows the code solution, but must run it to determine the actual result. This also clearly takes a while. At this point, this is not really a model anymore but a full agent able to use logic, trial and error to solve problems. This is definitely what AGI will have to be.

The Illusion of the illusion of thinking 😂

I asked my ChatGPT what she thought of Apple’s paper. She said “This from the people who brought you ‘Here’s what I found on the web for please turn off the lights.’” 🤣

9:35 “context limit” -> this is exactly why I prefer Gemini 2.5 pro, the 1M context window really make the difference.

You should ask o3-pro to write a program that will take the output codes it gave for the Towers of Hanoi solution and actually animate it, and prove that the solution works. I imagine it does, but that would be a double impressive test.

shared prompt for the 10 disk Tower of Hanoi: https://chatgpt.com/share/6848fff7-0080-8013-a032-e18c999dc371

Pliny interview! Get HYPE

There's only one metric I need, that's the asymmetry of Wes's eyebrows. This video smashed the record!

A part two of this video is necessary keep us updated please