We’re trying something new! Welcome to the first Endpoints Slack Interview, where we chat with interesting people in biotech, pharma, venture capital and science about what they’re working on, what they think and who they are.
Our first conversation is with Josh Wolfe, cofounder and managing partner of Lux Capital. He logged in to our top secret Slack channel last week, and the conversation has been very minimally edited for readability.
Andrew Dunn
welcome to the first Endpoints Slack interview!
Josh Wolfe
excited to be your guinea pig;)
Andrew Dunn
excited to dive into the biotech markets, AI, and more about your own life but I wanted to start with the bird app
Josh Wolfe
i am an obsessive user
Andrew Dunn
What’s your prognosis on Twitter and the Biotwitter community? As one of the more active users on there, curious to get your take on where we are in this weird moment for social media right now [Editor’s note: This conversation took place before Twitter was renamed “X,” and honestly who knows what happens now.]
Josh Wolfe
i think niche communities are more signal than noise. so lots of folks in biotech + techbio post science papers, debate drug significance, talk of FDA rulings, personnel departure… in essence sharing intel + insights and i see little noise within that talking about politics or kanye or trump or whatever. so i’m optimistic that niche groups filter out or reject noise. now general users… lots of complaints. but i’m optimistic (and hedged being also on Threads;)
and THANKFULLY little to no overlap with communities of crypto chaos…
Andrew Dunn
crypto twitter certainly feels like its own animal
i was about to ask on threads — what’s the status of the house a few weeks in? looking good? trashed? little boring still?
Josh Wolfe
yep, crypto mostly echo chambers of promotion + fomo.
on threads — so far lots of people cross-posting, i have a fraction of followers from Twitter, so engagement + comments are less kinetic. but the tone feels more civil. maybe that’s wishful thinking and we’ll all devolve into the worst of humanity! but so far better angels of our nature…
clear that both will coexist and i think Meta/Threads will eventually do promotion + exclusives (sort of what Elon is trying to do hosting people on Twitter Spaces)
Andrew Dunn
I admit I haven’t taken the time to set threads up yet, so you have to let me know if that’s another one of these twitter clones I need to try out. final social Q: do you think Twitter dies, particularly for that niche community of biotech experts?
Josh Wolfe
i don’t. i think it survives. with new CEO and Elon’s (rare) honesty of burning cash and huge loss of advertisers, and investors writing it down by 50% or more, in his own self-interest stepping back and just being Chief Tweeter and letting leadership focus on shipping better product experience will invite people back into the water. a lot of people left the platform because they resented his provocations + antagonisms + antics and voted with their feet (or fingers). a year from now i think it will be back—sort of like when everyone used to threaten deleting instagram or fb from their phones…and then come back:)
Andrew Dunn
heard it here first – chief twit lives!
alright, turning to biotech… i wanted to start with the vibe in the markets right now. i feel some solid tension and uncertainty between this prolonged pullback against real clinical wins, M&A is happening, even a few IPOs now.
let’s try this: 5 words you’d use to define where the industry is right now
love it — can you expand on optimism? curious on what’s driving that for you now
i also think what everyone wants to know is if the industry is out of the woods yet
Josh Wolfe
$XBI is about where its been on average last 7 years, of course down 50% from jan ’21 high; so the odds are reasonable it gets better not worse from here. general mood amongst people is rising as a lot of layoffs in biotech + pharma over past 18 months means there is a lot of available talent that was harder to get a year or two ago and in turn that new talent can join younger newer well-funded platforms. older platforms that were undercapitalized are shedding assets or mothballing/shutting them down. all of which is healthy. we’re seeing deals between pharma + VCs spinning out assets as newcos. so lots of consolidation + combinations which add to the optimism. and we’re seeing that scientific progress itself doesn’t halt even as greed + fear ebb & flow…and complemented with new computational and AI/ML techniques it is observably accelerating
is the “industry” out of the woods? maybe. but outstanding individual companies will be able to shine. rather than everything moving with high correlation, there will be higher highs for winners and lower lows for losers — in both public + privates.
Andrew Dunn
right, sounds like the haves and have-nots idea still very important. i’m surprised AI wasn’t one of your 5 words!
Josh Wolfe
obviously i felt covered as the letters were embedded in both “consolidAtIon” and “AlzheImers”
Andrew Dunn
well i stand corrected, twice
i did want to ask you for your thoughts on what Ramy Farid, Schrodinger’s CEO, told me earlier this year
he thinks AI hype is at dangerous, insane levels right now. And in biotech he said it’s pretty foolish for biotech startups to go touting AI as a central component, given much of the field is really using open-sourced machine-learning algorithms with tweaks for different uses
Key limitations are still quality data and actual scientific knowledge, he says
What do you make of those points, as someone who’s really closely involved in both AI and the biotech spaces?
Josh Wolfe
1. On his first point…
2. anytime there is high hype the cost of capital gets low. NVDA announced a $50M investment in Recursion and it added ~$1B of enterprise value.
and in some cases it’s deserved + awakening people to an important trend they have overlooked and in other cases maybe it’s investors jumping as they have into fads without focus on fundamentals…
i do think open-source foundation models are really only about to get started in biology. the context windows have heretofore been too small. but that is changing. i also think there will be small models with proprietary datasets (whether that is at scale of genetic, single cell sequencing, hospital systems, patient populations is still TBD). but the insights + analysis achievable at scale will shock people positively. relatedly companies like Gandeeva which have the ability to use cryoEM imagery giving atomic level precision and then predicting protein structures that will bind or disrupt whether in the case of PROTACs or molecular glues is a capability that was slow, serial and brute force in the past. so..YES there is hype. and lots of nonsense will be funded. and YES some of the most important new drugs will be discovered + designed computationally..and YES clinical trial design will still need human input and the humans themselves (even if patient population gets segmented more precisely) are still the ultimate laboratory…so YES i am way more optimistic that we are just getting started
but the work that the Meta/FAIR team led by Alex Rives did with ESM that beat Alphafold/DeepMind by say 2x and is now becoming what so many are using…is a preview of many small teams with lots of compute that are going to advance the field of drug discovery at evolutionary scale (even designing new cells and maybe organelles) we’ve never seen before…
Andrew Dunn
it certainly feels like that’s an inevitable part of the industry’s future. all these new biotechs with equal teams of software engineers/ML experts alongside wet labs and biologists and chemists
maybe one last bigger-picture Q and then i’ll close with some rapid fire Q’s on yourself, but on that point of needing human input and humans: the a16z team likes to semi-joke about the idea of Beach Biotech, or a startup that can be run remotely with automated experiments, robots in the lab, contractors handling manufacturing. How close are we to something like that being reality?
Josh Wolfe
even the composition of R&D leadership at big pharma is increasingly coming at problems with computer science approach
i also think there is an arrow of directional progress that points clearly to lab automation, science being done by machines, and the results being compared to old scientific literature using LLMs and then reverse prompting scientists to tweak parameters and run experiments (in the cloud) based on the results. scientists will be freed from the wet bench and way more focused on creative experimental design….
100%. i have said that we will be running experiments from a beach in the bahamas and you’ll be using your ipad to program assays that run remote just as you can use your ipad or laptop to obviate the need for studio time + instruments as a musician and can instead use protools/logic/garage band etc… i think within the next few years you will have maybe ~1% of global experiments and within a decade it will around 10%.
so it will take time, but once a few co’s show R&D productivity + happy staff + a competitive advantage…others will copy and it will gain momentum.
but it will be generational. younger scientist and those trained on these systems using Benchling + LatchBio and others will be way more early adopters than older scientists…
Andrew Dunn
anyone doing it yet?
or still just a few years out from being practical?
Josh Wolfe
Strateos is one and there a few others.. it is still early + risky but promising…
Andrew Dunn
That’s going to be a ton of fun to watch. Let’s close out with some rapid-fire Q’s on your own background, starting with growing up on Coney Island. How’d your upbringing make you different, particularly thinking of so many in the industry probably growing up in Manhattan or the burbs?
Josh Wolfe
single mom, she worked multiple jobs, we had no money, 4 of us (my grandpa who delivered the NY Daily News at night + grandma a retired ‘meter maid’) lived in 2 bedroom 1 bath apartment. everyone in my neighborhood had some hustle or con. i grew up skeptical and distrusting assuming everyone was trying to screw us over or scheme. so i grew up squinty-eyed assuming the worst. i have a line everyone at Lux has been indoctrinated with: FAILURE comes from a FAILURE to imagine FAILURE… it means expect the worst and imagine risks and figure out how to kill those risks. it’s as much derived from a psychological protective mechanism as much as anything else…
everyone i knew from manhattan or the burbs…had money. and grew up with silver spoon. i was deeply resentful + motivated. i like to say CHIPS on shoulders put CHIPS in pockets—and its a trait i look for in our founders…people with something to prove
often no matter how much money or success they achieve—that fire still burns.
and its where societal progress comes from. lots and lots of disaffected people who want to prove others wrong + live a better life. and we all benefit from that!
Andrew Dunn
not many VCs I know that preach expecting the worst and imagining failure
rapid fire #2: What pitch do you most regret passing on?
Josh Wolfe
we saw Maven Clinic at the seed and A and passed and ended up paying a much higher price later on to partner with Kate! that was an expensive miss.
adjacent in our core tech investing we passed on disciplined price setting on Cruise Automation — and during diligence my partner Shahin had intro’d them to GM. another fund offered 80M vs our 40M pre. a year later GM bought them for a $1B. would have been 11x in under a year. ouch.
Andrew Dunn
alright, two more rapid fire Qs to close: link to a scientific paper (or preprint) you’re obsessed with right now
Evolutionary-scale prediction of atomic-level protein structure with a language model
“Lin et al. trained transformer protein language models with up to 15 billion parameters on experimental and high-quality predicted structures and found that information about atomic-level structure emerged in the model as it was scaled up. They created ESMFold, a sequence-to-structure predictor that is nearly as accurate as alignment-based methods and considerably faster. The increased speed permitted the generation of a database, the ESM Metagenomic Atlas, containing more than 600 million metagenomic proteins…”
Andrew Dunn
Rives! I was blown away by AF2, so yeah that’s hard to even fathom
Josh Wolfe
Alex helped us start Kallyope and i’m a big fan. and a big fan of helping him build whatever he might do next…;)
Andrew Dunn
Final one, maybe balancing out sharing your biggest miss: what startup are you most excited about right now?
Josh Wolfe
hard to choose between any of our favorite portfolio co’s––but 2 are Osmo.ai (digitizing smell) + EikonTx (Roger Perlmutter at the helm + Nobel work of Betzig) is amazing combo.
this was great fun! look fwd to our next!
Andrew Dunn
Thank you for hopping in our Slack! Who should we invite next?
