Mike Israetel

Sure, sure. The solutions to the problems that we're seeking... to systems that intelligent, should they choose to solve them, can be, for lack of a better term, pedestrian in nature. And they're going to be dealing with problems that are much more complex than the reengineering of human biology.

So for me, when the raw compute and the raw understanding of how to manipulate matter and energy to get kind of any kind of shape you want at a given energy input, When that's there, the only question is like, are we going to try to do it or not? And that's where I come back to the incentives and constraints problem.

So for me, when the raw compute and the raw understanding of how to manipulate matter and energy to get kind of any kind of shape you want at a given energy input, When that's there, the only question is like, are we going to try to do it or not? And that's where I come back to the incentives and constraints problem.

The biggest hurdle to the development of advanced pharmacology and genetic engineering and so on to do this kind of thing is going to be regulatory in nature, hands down. FDA, everything's off by five or 10 years. It sucks. But once AI has enough time to cook on these problems, the candidate drugs released will run through trials with just an unreal record. But why?

The biggest hurdle to the development of advanced pharmacology and genetic engineering and so on to do this kind of thing is going to be regulatory in nature, hands down. FDA, everything's off by five or 10 years. It sucks. But once AI has enough time to cook on these problems, the candidate drugs released will run through trials with just an unreal record. But why?

Because if you have very not so good at things AI, that's decent.

Because if you have very not so good at things AI, that's decent.

It doesn't streamline it at all. It just flies through it. Like knocks out phase one, knocks out phase two, knocks out phase three market. So you can say-

It doesn't streamline it at all. It just flies through it. Like knocks out phase one, knocks out phase two, knocks out phase three market. So you can say-

But phase one to phase two to phase three, it's still going to take a decade. Totally. But at the end of that decade, we have super drugs hitting the market all at the same time, as opposed to the incremental process. The increments are all handled upfront by the AI. And that last decade is just like, we just got to do this.

But phase one to phase two to phase three, it's still going to take a decade. Totally. But at the end of that decade, we have super drugs hitting the market all at the same time, as opposed to the incremental process. The increments are all handled upfront by the AI. And that last decade is just like, we just got to do this.

Yep. We already knew how to build retatrutide back then, and we could have just done it. No one cared because the money wasn't there. Slash, there's lots of other candidate drugs you could work on. That's interesting. Yeah.

Yep. We already knew how to build retatrutide back then, and we could have just done it. No one cared because the money wasn't there. Slash, there's lots of other candidate drugs you could work on. That's interesting. Yeah.

Yeah, and so if the AI is powerful enough, it'll just give you candidates that are just killers right offhand.

Yeah, and so if the AI is powerful enough, it'll just give you candidates that are just killers right offhand.

Eventually all of it. But I'll give you the second rung of what's starting to happen now. The second rung, the first rung is like candidate drugs based on protein structure alone. And will that protein structure fold into the receptor we're targeting well enough to give us some activity? The second phase is... This sounds funny to say, but it's computationally going to be tractable quite soon.

Eventually all of it. But I'll give you the second rung of what's starting to happen now. The second rung, the first rung is like candidate drugs based on protein structure alone. And will that protein structure fold into the receptor we're targeting well enough to give us some activity? The second phase is... This sounds funny to say, but it's computationally going to be tractable quite soon.

Simulating every single protein in the human body and seeing how that candidate drug interacts with every single other protein. And then you just optimize the selection criteria for- Dial up the effect, dial down the side effect. Are you familiar with Jepperon, aka Exua, a new major depressive disorder medication? No. So Jepperon, the trade name is Exua. What class? Targeting SSRI, I think.

Simulating every single protein in the human body and seeing how that candidate drug interacts with every single other protein. And then you just optimize the selection criteria for- Dial up the effect, dial down the side effect. Are you familiar with Jepperon, aka Exua, a new major depressive disorder medication? No. So Jepperon, the trade name is Exua. What class? Targeting SSRI, I think.

And it only targets the serotonin receptors in specific parts of the brain as opposed to just like, you're going to get it all. And so it has seemingly no more probability to reduce sex drive or alter consumption of food patterns than a placebo. That's not even developed with AI. That's just a more selective targeting. we can get almost 100 to zero ratio targeting with that phase two approach.