masterjack

> Please explain in your own words why you believe this is not the case.

It wasn’t 2023: Last post 11 months ago, last comment 8 months ago, which is a typical level of lurking

I thought they were projecting 10B and said a few months ago they have already grown from a 1B to 4B run rate

Remarkably there’s really just one way to tie them together, you can always manipulate the knot to move between the different variants

13/183=0.071 so how can it be 20pct for this round??

I may agree if it was a 20% dilution round, but not if they are increasing from 3% to 7% dilution. Being so massively oversubscribed is a bullish sign, bad companies would be struggling to fill out their round.

It will typically be the same 40 GB model loaded in, but called with many different inputs simultaneously

Yes, you can `print axioms` to make sure no axioms were added, make sure it compiles with no warnings or errors. There’s also a SafeVerify utility that checks more thoroughly and catches some tricks that RL systems have found

It’s possible that in the Google deal you had to agree to sell back the shares (at a low value like par or original strike price) and the 1% refers to either those proceeds or the size of the Google employment package. If you didn’t agree then you would be left holding your shares of a company that is now gutted.

Exactly. And presumably had a more sophisticated harness around the model, longer reasoning chains, best of N, self judging, etc