They use Neural ODEs as the denoiser, using multiple neural ODEs in a chsin and they somehow stick a time emedding into them.

No, it isn’t. It’s perfectly comprehensible.

The description of the architecture isn’t incredibly clear, but it’s enough to get the idea. I’d have liked to see the details, but if they want to write it like this that’s fine.

It’s a real firm. They’re associated with Ycombinator.

The direction of achieving robust reasoning in LLMs is both reasonable and likely to be achievable. It probably isn’t super easy, but lots of people are thinking about ways of achieving it and it’s an active area of research.

I’m not sure whether I’d invest in it, but it seems fine. It’s a reasonable direction and I’m sure there’s a way for people to be productive while working in that direction.