Angels, labs and biosecurity

Week two is over. We explored the wet lab options around SF, and seem to have decided on where to land. Originally, I was trying to avoid BSL2 (biosafety level 2) to make things faster and easier. We toured a BSL1 lab before in SF, and it was great, except for the lack of BSL2, and we couldn’t proceed with them.

Why? For our PoC we would anyways need to work with human cells, and that’s basically landing us in the BSL2 category with only 1-2 reasonable labspace options. The only annoying thing is that they are slow as hell and keep on scheduling next steps with a weekly cadence. We have already pitched our project to a committee (7 people!) of one of them, and got accepted. Today we’re touring the lab space, and, hopefully, will be able to move in and start our wet-lab journey soon! I was thinking that this option is snail-paced but the other one scheduled to talk to us only on September 22nd — it seems like they are very busy.

We’ve picked a list of four potential biological problems (which currently seem impossible to solve or extremely challenging). We want to start tackling them in the initial PoC stage. If you think that four problems are too many, and we need to be more focused — we agree 100%. That’s why we picked 4 problems which are orthogonal but can be solved within the same stack of tools, reagents, AI engine, etc. TL;DR — parallelization is a feature, hence some apparent “lack of focus” and four problems running in parallel because our product is not each of those solved problems per se but an engine that can prove that it’s able to do all of that cheaply and in a horizontally scalable way.

All of it is by design within a limited time and budget constraints. Our thinking is that whatever we would be able to solve within the first 3-6 months, and the budget we would spend in doing so are going to provide a ballpark estimate for how much money and resources we would need for our future scientific superintelligence with our wet lab in the loop to solve one problem within a horizontally scalable platform.

One of our North star downstream goals is to be able to push 100 new biologics at the same time horizontally from a fancy target or a peculiar piece of biological evidence to a potential drug candidate stage with some convincing in-house data showing that it’s working for a relevant application.

Why 100? Because to be able to tackle aging at some point, we’ll have to tackle a lot of orthogonal mechanisms at once — people say there are 300 theories of aging. Ideally, when we succeed, we will be able to test all of them horizontally within the same 3-6 months and our current PoC budget times 100. My bet is that we will be able to do it well within the budgets previously raised by the prior generation of “aging” companies, e.g. Retro, NewLimit, Calico and Altos Labs. It’s a bet — we’ll touch base in half a year from now and reconsider.

TL;DR on the execution side last week: we managed to process almost all the initial inbound requests to chat (>10 angels and VCs), the incoming requests are sprouting, and we’ve decided to be smarter onward by keeping a more realistic cadence of calls. Some cool people are already willing to join our team, and we’re trying to find a good way to process these requests. In addition to the physical reality bottlenecks (like trying to rent a fitting labspace), one new discovery on our superintelligence journey is that both Anthropic and OpenAI do not want us to solve biology and develop drugs. We were thinking about smart reasons why big labs wouldn’t solve biology for us after having claimed to solve maths. Reality is way more boring of course, and the two reasons came straight out of their own tools.

I don’t believe they will solve biology if they are not able to design their own guardrails smart enough to distinguish drug development to cure multiple diseases and, hopefully, aging from developing bioweapons. If I were them, I’d fire their whole biosafety teams and hire someone more competent. It’s an interesting world we are living in.

Share on: