And so they're still working on it.

Yeah.

And like when we think about like additional levels of training, there's like kind of a bidirectional process that occurs once a model is formally launched for like public consumption.

You know, a lot of people don't realize that when they consent to use one of those models, it's going to automatically take whatever you give it and take it back to kind of like the source model.

and continue to shift and change its algorithms based on like hundreds of thousands of user input.

And then you have like kind of a microcosm of that level of learning and shifting when you are just engaging with the platform as your own user.

And so even if you have what we call like an enterprise agreement, which kind of shields you from a lot of that sending your data back to kind of the larger server, you're still going to see that gradual evolution.

I rarely use PubMed or other search tools that I used to use before.