Disclaimer: Thoughts are my own. At the time of writing this, Gordian is still around and doing very well. But startups grow fast so the Gordian I knew then may not be the company that exists today, so the Gordian I will write about will be referred to it in the past tense.
Gordian is a pure biotech. Over the years it branded itself in many ways; longevity biotech, in vivo perturb-seq, AI-first biotech, etc. These descriptions are all true, but at its core, to us nerds, it was a biotechnology company that wanted to be the next Genentech.
A multi-decade play, tackling some of the most debilitating, chronic, expensive diseases of our time. We raised our Series Seed around 2019 to build our platform, then our Series A in 2021 to scale and apply it. For most traditional biotechs, especially of the past few years, that’s almost unheard of. Though arguably, at least for AI companies, venture timescales are now once again reaching for earlier and earlier stage companies, re-fueling the spirit of deeptech startups.
Matin and Francisco.
Drug discovery is a test of endurance. There are few “get rich quick” schemes, and tacit knowledge is fun to say, but hard to find. You need good leaders to make it through. Martin (CSO) and Francisco (CEO) were some of the most passionate, intelligent, and devoted people I’ve met. They’re those leaders who are working just as hard as you, even after you did another 80hr week. It’s just hard to take days off when you see people like that. Aside from family and working out, I rarely saw them not giving their 100%. At least from the outside, they were always present, asking how the team was feeling, and intensely focused on the science. That passion was infectious, and I still try to exemplify that devotion to my work to this day.
They gave back to the community. Martin, an intensely passionate geroscientist, sponsored AGE conferences, mentored young scientists, spoke at events (where mutually beneficial), and always made time for my musings. He recently even gave a TED talk in Cambridge, if you want to check it out. Francisco has lore. He grew up in California, was a child prodigy, got a Stanford physics PhD, blew up batteries at early Tesla, serial founder, the whole gambit. He’s obviously smart, which I respected, but the first thing I always heard about him from the community was his kindness, devotion to his family, and, for lack of a better description, a “dadness” that certainly gave him wisdom and grey hairs.
I write about their positive aspects not because they were perfect, but because that’s what they tried to exemplify for the team and to others. I don’t respect them because they were perfect, because they almost never were. But what I know for certain: that naïve kid from the Japanese countryside wouldn’t know what he knows now, writing this from London, if it wasn’t for them showing me the way. I can’t be a child prodigy anymore. I won’t be friends with Laura Deming since I was 16, receive a K99 fellowship, or blow up those OG Tesla batteries. But I know that there’s a role for me out there to play.
The greatest lesson here: companies are the physical manifestations of the founders’ personalities. The culture, the output, the people, the reputation, is mostly steered by the founders themselves. Just as not all companies are created equal, neither are founders. Of all the lessons I learned, the most valuable one was how to identify great leaders and founders. Throughout the years, I’ve been fortunate to meet many great founders, but also many mediocre ones. The Ghost of the Valley, as ____ and I say. Those who aren’t devoted to their work, their craft. Those who may have had a good idea once before, perhaps at 20, when they became a Thiel fellow. Those who might be effective fundraisers or like their cool technology, but are not completely devoted and desperate to solve civilisation scale problems.
Talent
Gordian had one of the highest talent densities I’ve ever seen. It gave me an incredibly high standard of excellence, sharpened my “radar” for talent, and set a benchmark for how science should be done and experiments executed at a specific speed. There’s just something magical about finding different people who are the best at what they do and putting them all together in one room with one mission. Intellectual freedom is limited by the resources that enable it, and startups with great people are one of the best places to do groundbreaking work at breakneck speed. I learned a lot of domain tacit knowledge, but what was most valuable was the peripheral knowledge. It’s the stuff that Richard Hamming talks about: sitting next to the molecular biologist at lunch after a long day of surgeries, asking them about random topics like Golden Gate assembly or Sleeping Beauty transposons. The rate of learning was truly insane.
Hiring good talent has a multiplicative effect that’s hard to predict. Much of Gordian’s success in building a bespoke platform came from its obsession with talent. But just like how we’re bad at predicting LLM improvement rates nowadays, it’s hard to overstate how important the founding team is, not just for raw scientific output, but for setting the culture right to encourage failure and ambitious thinking. To this day, the absence of a true “fail upward” mindset is one of the biggest constraints on ambition, narrowing the range of bold bets and high variance swings we would otherwise be willing to take. Currently, I’ve only seen this mindset consistently cultivated in ambitious deep tech startups.
Good talent is hard and expensive to find. Sometimes you put up a bounty, find personal phone numbers on Grok and cold call, read PhD dissertations that probably nobody else even opened. You go to the ends of the world to fly out to meet them, to show that you’re serious. You shoot down the bird with a bazooka because overkill is the expectation for great talent. Because if you think they’re really that good, it’s only a matter of time until someone else takes them. Being obsessed is the only way. Poach people, within reason, because the best people are already doing their best work for someone else, if not for themselves.
Predictive validity
Many biotechs, as well as VCs, fundamentally do not understand (or perhaps sufficiently value) predictive validity.
Coined by Jack Scannell, the concept points to the root failure mode of many biotech companies. No matter how fancy or sexy a new platform, modality, or AI model is, they’re fundamentally just tools to find a drug. After finding that drug, you still need to test it in a cell line, mouse, or an organ on a chip model before testing in humans. That’s still where most biotechs fall apart.
They either 1) see negative data in their dog or primate study, or 2) see false positive results in their large animal study, and the drug fails in patients. This is an expensive mistake, but what’s understated is how heartbreaking it is for the patients who participate in these trials. Patients who are desperate for a cure participate in risky trials to try an experimental drug because of hope. As drug hunters, we have an obligation to take the predictive validity of our experimental models seriously, and not get caught up in the allure of an all-knowing AI oracle.
Predictive validity was at the centre of Gordian’s thesis. During my ~4yrs there, people always remembered us as the company that was testing on horse knees, specifically retired racehorses as a naturally developing OA model. If OA, for example, predominantly affected old, obese women, we would try to use old, obese, female mice in our studies. Same with HFpEF: if the classical clinical phenotype was an old, obese woman who smoked a pack a day, we emphasised using older, obese mice with long congestion where possible. As an investor and advisor now, I rarely see the same attention to detail in other biotechs, especially those run by AI scientists or first time founders. With more AI forward companies stating that they will “predict” biology better, I’m curious to see how many of them actually pull it off. Only time will tell, but the cynical biologist in me thinks they will likely fail.