So when we think about AI, AI is really a question about data. The more data that you have, the more you can learn from. So when we apply this to something like language, where you have an alphabet of 26 letters, these 26 letters can be put together into millions and millions of words, and then those words can be put together into language. And so AI is really good about collecting all of this data and then using that to now predict new phrases, new responses. There's a really good similarity to biology, where biology also has letters that make up information. And so you have 20 amino acids. These are the letters of biology that make up proteins and enzymes. And so you can put those letters, those amino acids together into millions of different ways that make different types of proteins, all based off of those same letters. They all have different functions. They all have different functions. So the more that you can now learn what those different functions are, you can now start to predict different properties such as stability, activity, substrate specificity, which all contribute to your ability to learn and design. And so what's lacking is the readout. And so this ability to not just predict, we all know what a word is and how it should sound and what the definition is. We don't always have that information for, a protein or an enzyme. So while you can predict what the protein letters might be and maybe what its structure looks like, you don't always know what the function is. And so we can solve that gap using the tools that we've developed to really get that function data quicker. And then we can learn by making changes, very specific changes in the letters of those amino acids. And we can quickly iterate to say, okay, this specific change now had this specific effect on function. And once we know that, and we can collect enough data with enough variance, we can then learn from that. AI is a fundamental tool that we use to really speed up timelines. So this ability to engineer enzymes, it's complex, it can take a lot of time. But the ability now to predict faster, allows you to test faster, allows us now to develop really focused on the libraries very quickly. That allows us to really short circuit and speed up that process of generating libraries and screening them. And so we don't have to pass our sequences through an organism, we don't have to do sequencing, we get all of the data right out up front. And so it's an essential tool that allows us to design, build, test faster to develop now more functional enzymes that can be applied to biomanufacturing, biocatalysis to generate those new molecules. And so that's what we're trying to develop. So the way we fill this gap is because we fill this gap is because we can actually measure those functions. That's the missing link. That's the data that's needed to then start learning. So I think what you're going to start seeing over the next, you know, couple years is we're going to be able to generate more data specifically for the enzymes that we're interested in. And so we're going to now be much better and faster at engineering those functions that we're going to be able to generate more data specifically for the enzymes that we're interested in. And so we're going to be able to generate more data that we need to generate more data. And so we're going to be able to generate more data. We're going to need more data and be able to generate more data. And so that's the key here. I think that's where some of our secret power lies is this ability to generate the data to refine the models to then continue to learn using these large language models. I think that is applied to specific enzymes for specific functions. I think if we're looking forward, you know, on the multi-year timescale, I think the ability to now link enzyme steps together to now predict series and cascades of reactions, I think that will get even better as well, because now you can couple biological pathways with specific function of different enzymes, and that allows you to now develop systems that work much better to get from some low-cost substrate or input much more efficiently to your product of interest, be it, you know, a nutraceutical, pharmaceutical, or specialty chemical, whatever it is.