[alecco]

> They know Google has a ton of data to train LLMs on.

And they have a massive amounts of TPUs. And yet... their models are way behind.