The first month's summary

It’s been a month since I left Retro, and here I am — I’m able to run any experiments at any scale in our own lab, using our proprietary AI engine, having raised enough money for our wetlab PoC (or rather 4-5 PoCs). Mischief managed!

It’s been a month since we started. It may be the right time to summarize my general impression. “Do not solve problems that haven’t emerged yet” is definitely the motto. I barely have any time to respond to people or read thoroughly.

We launched our wetlab on Monday, unleashed our AI engine to design a few pilot sequences for debugging our workflow, and the first DNA order has already been submitted! The first order is usually the hardest — from now on, it’s going to be easier. I’ve spent a ton of time thinking about a gazillion tiny things we need to do in order to become fully operational. The good news — actually, great news — is that the lab we’re renting has some crazy advanced equipment that made me completely rethink our wetlab approach. I believe we can now do 3-10 times more than I thought we’d be able to do.

Our fundraise, which we officially haven’t even started yet, has already brought in enough money for me not to think about money constraints while setting up the lab, which is a nice world to live in!

Although we haven’t officially started fundraising yet, we have been getting a lot of requests from angels and VCs to chat. I have an intuition which I cannot fully describe, but I can feel whether someone would invest or not. So far the hit rate has been very high. Fundraising is like dating. I am able to read the room the moment I walk in. It’s immediately clear whether these guys are going to invest, or they simply invited us to chat and pitch out of pure entomological interest. In the latter case, we simply get an opportunity to practice our pitch once again, and get more feedback. In any case, it’s a win-win situation.

We’re learning fast from iterations, and use every opportunity to run more of them. We didn’t have a pitch deck until two days ago. A good and big fund insisted on us presenting, even though we’re not fundraising yet. We did present, and prepared a pitch deck for it. Now we have one, and I already like it. I like it when we go from 0 to 1 along any of the infinite axes that we have to optimize on our way to success. Now, we’re one step closer.

It’s been my first week in our own wetlab. TL;DR — we’ve already started the first spin of our lab-in-the-loop AI flywheel for protein engineering, and it’s not even the end of week 1. I’ve been in a very fulfilled and content state, and I haven’t felt at ease to such a degree for a while. Of course, it’s been challenging — not in the sense that I don’t know what to do and get lost. It’s quite the opposite: I’ve started assembling a gigantic puzzle piece by piece. The first two days I couldn’t really order anything because I had to send petitions to every single vendor explaining who we are, where we are and what we are planning to do, and get their approval for releasing any shipments to us. On Tuesday night, Pandora’s box finally opened — and I could order the first set of our AI-engineered proteins. For now, it’s a small-scale pilot to debug production, purification and some basic assays. I was allowed to order cells on Wednesday. On Thursday, I received them, and started expanding them! I had to borrow a cell culture flask from a neighbor to save a week downstream while waiting for our own order of those.

It’s a shared lab, so the tissue culture hoods are basically empty — and the locals drag a tiny little cart to transport all their pipettes, tubes and reagents to and from the tissue culture room to their desk in the open space. The same recipe was invented by homeless people in SF, who carry all their baggage along around the city. It’s a minor inconvenience though, and, interestingly, the throughput and efficiency of using the lab space in general are extremely high as a result — there are almost 100 independent companies working in a space footprint smaller than the one we had at Retro.

Being a single scientist in a wetlab made me feel like a single father, or a solo traveler. If you’ve ever tried either of those, you know how even the most trivial thing becomes challenging. Like going to a restroom at a café with your luggage — you either clumsily bring it along, or risk losing your documents along with a stolen bag. I do miss having my former peers around to cheer each other up with cryptic memes that nobody else could understand when biology turned out more complex than we had anticipated, and the world we lived in turned out to be less than ideal. It’s a temporary state, of course — but it’s the only thing I actually miss from the last few years at Retro.

I’m turning into a frontier lab doomer — for most bio work I’ve switched to open models, and they work great! I don’t know how Anthropic and OpenAI are going to justify their valuations when an open model beats them on all the practically relevant work. What’s their moat right now, if it’s not the models? It’s started to remind me of the dot-com bubble — imagine having “creating websites” as your moat in a world where everyone can make websites.

It has always been true in bio that models are never a moat, and now the same is coming back to bite the founding fathers of general-purpose LLMs. Prep your popcorn — the bursting bubble is going to be painful for them! And extremely valuable for everyone else who will use democratized and commoditized LLMs for their work (including us!).

Share on: