Beth Lyons

Okay.

And DeepThink Gemini came out with a high score and they did it, I think, for trying to find the money.

Poetic did it for thirty seven dollars per answer.

And that's significantly lower than whatever DeepThink was was charging or how much it cost for DeepThink to do it.

And that is exciting because, first of all, it's open source.

The repo is in the announcement.

You can go and load it yourself, put your own API keys in and see what kind of results you get.

And also that it is a meta system for reasoning and that more and more we're seeing that it's not that you have to just shove more gazillions of dollars into training.

to get better results.

Now we have some stuff on the back end, which involves less hoovering of data or creating synthetic data, less energy, because the inference energy so far still is lower than the training energy.

That will flip maybe as usage goes up, but

These are like good, good results or good indications that we're going to be able to get higher levels of AI without needing to do that order of magnitude thing.

Right.

Which was 10,000, 100,000, a million, 10 million.

Right.

We were just looking at those zeros in training and getting better and better results.

And the announcement being, we wipe the floor of you with you.

Yes.

Right?

Like, look out, Google, because we're coming after you release.