 If you want to build an AI model using deep learning or the latest science, it's typically about a month or so to kind of define what the problem is, well, to make it an AI problem. Takes you another three to six months to gather the data, another three to six months to build the model, and then you ship it. That's six to 12 months. Now you can pull one of these foundation models, like BARD or ChadGPT, and give it a couple examples. That's like an afternoon. And then you can take the API and integrate it within a few days. So you're going from a year to a few days for a practitioner who may not even know it that well. That's amazing. And that's, that all hit us six months. And so you're going to see this proliferation of people, you know, AIs, classifying cancers, classifying everything else. The tools now are getting like a spreadsheet. So powerful. People have an idea. I got an idea. Here's the data. It's the phenome. I can have a result. I can publish it. Boom. I've got a classifier that used to take a year I did within a week running in a cloud. We haven't just begun to see the impact of that. Sanjay.