Microsoft's Shital Shah argues Meta's reported shift to closed-source personal assistants misses the true purpose of superintelligence

@teortaxesTex I think they don’t want their models being independently benchmarked lol

@sytelus Amen. Baffling they don't understand that simple thing. Most of the world's generated tokens will come from commodity models; I would even go as far as saying most tokens will come from on-device models, far away from the frontier

@thewhiteboxAI A 100T model will give approx same outcome as 10T or even 1T model if you were designing usual web app. Major value of super intelligent will lie in dramatic developments like battery that can last 10X longer or gene therapy that can cure cancer.

@sytelus Agreed.

Been there. When you've got a genuinely strong model, you're begging for independent benchmarks — they're free marketing. When you're nursing gaps, every external eval is a liability. The shift from 'look at our internal scores' to 'benchmarks don't matter anymore' is the most predictable move. Every time.