tl;dr: software got its dorm room moment. biotech hasn't. here's why that needs to change, and three ways to start.
one of my favorite things to do with my friends in Startup Shell is throw around the myriad startup ideas we have. what happens after we stumble onto a good one, however, is quite different.
i generally end up calculating the cost to run the most basic wet lab proof-of-concept experiment, realize building an MVP worthy of angel investment will take 6-12 months and 100 times more money than i have currently available to me, and file the idea away under my ever-growing list of potential phd projects or things to fund when i win the lotto.
what my software friends can do with their good ideas, on the other hand, is quite different.
over the course of a few days, from their dorm room, shell, the classroom, wherever they want, they'll build a fully-functioning MVP. a few weeks, they'll have double-digit customers. a few months, they'll have traction, get into YC, or abandon the idea - to start the process all over again.
i point this out not to say that the software/SaaS startup cycle is fickle. these rapid build-launch-fail-repeat cycles are actually, in my opinion, one of the most impactful phenomena in the world of tech. failing fast is building younger, more resilient, more experienced, more creative founders. the impact is noticeable - total enterprise software VC jumped 64% last year. of course, not all of this can be attributed to the fact that tech founders are getting younger - but after all, younger people are naive enough to challenge existing paradigms and pursue novel solutions to problems others may see as norms.
put another way - the democratization of software development, there from inception but now infinitely enabled by generative AI, has led the greatest proliferation of creative software solutions in history, and to an incredible phenomenon:
a student can now build a billion dollar software company in their dorm in a month.
this is wonderful, of course. but it also begs the question: where is this moment for biotech, and for young biotech founders especially?
it's an important question to answer now of all times, for many reasons. for one, the problems biotech needs to solve sure aren't getting smaller.
but there's an interesting argument to be made on the economic side too. many voices in the software community, including ones i talk to daily, have started to point out that you can increasingly one-shot many popular SaaS companies in, say, cursor or claude code. a competitor has a feature you like? plug it into an LLM and ship it in a day. some companies built around popular large language models have seen their entire concept released as a chatgpt feature 2 weeks after launching, killing their entire business. by and large, large language models are becoming increasingly generalizable for software solutions.
given this, one could argue that the SaaS industry is at risk of cannibalizing itself in the next couple of years due to AI-enabled oversaturation and overcompetition. if that happens, where will the investment money pouring into SaaS go? entire firms who are heavily invested may just go under. many more will be looking to pivot. if that money can't find an outlet, it could have dire consequences for the productivity of the startup economy.
SaaS going under could lead to increased interest in funding biotech. but biotech itself isn't ready to meet the demands of investors who demand rapid iteration and endless ideas. it's slow, governed by credentialism and insane startup costs. younger founders, so in demand in other sectors, find themselves unable to get off the ground. biotech just simply isn't democratized in the same way software is.
even if you disagree with my assessment on SaaS, it's hard to argue that biotech isn't overdue for an honest conversation with itself on democratically increasing accessibility and buildability. if we want to enable a golden age of bio-innovation biotech needs its dorm room moment.
below, i outline 3 ideas to bring about this moment and to platform younger founders. the central thesis is simple: young software founders face very few barriers to derisking their business hypotheses. we need to lower barriers for young biotech founders to do the same, and get them excited about doing it.
first: we need to be honest about what our computational tools can and can't do, as they are a critical on-ramp to democratization.
computation in biology has come a long way. during my time at nextcure, i watched binder generation software go from 10% hit rates to 50% and beyond. everyone has heard of alphafold - alphafold 3 in particular, with its ability to simulate protein-dna complexes, is enabling us to make faster and better inferences about biological structures and how to engineer them. it's important for us to acknowledge these advances - computation is faster and infinitely more accessible than the wet lab. it's also important to realize that gen Z and gen ⍺ grew up software-literate. especially for young founders, software is essential for democratization.
at the same time, many of our tools are hard to find and harder to benchmark. we've reached the point where models and tools are going straight to biorxiv, mirroring the speed with which traditional software is developed and released. unlike traditional software, though, we don't have defined benchmarks, which makes it more difficult for non-domain experts especially to locate the optimal tool for their usecase. however, groups like rowan and tamarind bio have begun to fill this niche, which is an encouraging signal.
a larger problem is that many tools are incomplete. alphafold works wonderfully for many classes of proteins, but it's missing coverage for many more. generative softwares still rely on manual docking procedures subject to human error. protein and dna language and structure models can't quite capture all the intricacies we observe in nature. LLMs still struggle with understanding biological truth in the same way they can digest software paradigms. a majority of biological phenomena don't even have software models available.
second: until we have more robust foundational biological models, software cannot be the entire solution to democratization. for the phenomena we can't simulate, we need flexible, cheap, and accessible infrastructure to bootstrap hypotheses and build community around.
testing a software hypothesis is quite easy. programming languages are highly intuitive and can be implemented straight in the console or in a notepad app - or generative AI can write code for you. our laptops can handle initial tests on-the-go, and compute for more strenuous tasks has gotten outrageously cheap (memory shortage notwithstanding). biotech is the exact opposite. lab access and reagents are expensive and limited. designing protocols requires synthesizing papers across decades and standards across manufacturers. one tiny mistake can cost you weeks. and sadly, you can't drag a biosafety hood into a dorm.
one could immediately suggest that college-aged founders or founders in academia utilize their university's resources. oftentimes, this is not possible or desirable for several reasons. university tech transfer offices can negotiate quite aggressively for rights over any invention made using university resources. university-based tech founders can skirt this quite easily by running local simulations or purchasing cheap cloud access. we biotech founders cannot. space at universities is at a premium and is thus quite expensive. young founders especially face credentialism as a barrier when trying to find traditional support through universities.
solutions to these problems lie in democratized, roboticized infrastructure. i'm not an automation-only evangelist - i still think that learning to handle a pipette is valuable, and running protocols by hand can highlight inefficiencies that would otherwise be hard to spot. however, simplifying protocols down to the simplicity of software API calls is doable via automation, and immediately makes the lab more accessible. biotech needs cloud experiments the same way tech has cloud compute.
imagine designing a protocol with a software copilot, deploying it, and then going back to enjoying your morning. having time to read papers and meet collaborators while your experiments run themselves. designing autonomous flywheels that go back and review long-standing procedures to determine if they can be sped up or optimized, saving even more time. being able to schedule novel research and analyze your data around your class schedule. cloud labs enable that for biotech, and allow for us to fill in the gaps that our current computational softwares cannot address.
the industry seems to also agree that cloud labs are the way to go, at least in some form - we've got a few floating around, with ginkgo being the most notable recent entry in the space. the problem is that protocols are still quite limited, throughput quite low, and prices quite high. while more users flocking to cloud labs would probably drive prices down, it's not exactly fair to expect people to pay an early adopter tax. meaningful price decreases will probably come as the surrounding instrumentation gets cheaper. trilobio is currently leading the field in this charge.
another solution to the priciness of centralized cloud labs lies in decentralization and the proliferation of community labs. community labs already come with some major benefits, the biggest being that they provide a shared space for builders to meet and share ideas. community labs are already hacking together affordable lab automation solutions (as far as i know, biopunk has an active lab automation project), and may be able to provide researcher-specific flexibility that larger providers cannot. as a sort of symbiotic benefit, automating lab equipment may help mitigate many safety concerns that local and regional authorities have concerning ""hobbyist"" biotech. however, funding community labs still remains a concern - one that needs answers besides "more membership fees".
ultimately, cloud and community lab infrastructure offers a path to novel dorm-room science the same way cloud compute enables dorm-room foundation model building.
third: more universities and k-12 institutions need to start inspiring young people to build in their dorm rooms in the first place.
growing up, i was surrounded by programs that encouraged learning about software. lego mindstorms. girls who code. scratch workshops. FIRST robotics. new comp sci classes. usaco. hackathons. a specific, coordinated effort to not just get students to know what software was, but to get them to enjoy creatively putting it together to compete and solve problems.
that effort worked quite well. too well, actually. computer science is now one of the most saturated fields in the current job market. but as demand has slowed for software engineers, programs that excite students about other sectors like biotech have not picked up at all. while there are a couple programs that try to market towards high schoolers (biobits and other kits are relatively accessible even if they only focus on very specific experiments, and high school iGEM is alright), no one is marketing to younger students, and programs across K-12 and college need more focus.
further platforming existing programs would be a great start. iGEM especially provides a great example of what developing a biotech startup looks like. more support needs to go towards educators who want to start high school iGEM teams. one huge way to do that could be encouraging university iGEMers to start and/or mentor high school teams, either through a community initiative or through incentives in the competition itself. at the collegiate level, US-based teams especially need to coordinate and negotiate more funding from their host universities.
platforming and supporting local community biolabs if they exist near you, or founding your own if one doesn't, would also be a great start. more biolabs means more space for biotech community to coalesce and get inspired. more volunteers at biolabs means more people who can go into the community, host community events, and help mentor students through their first biotech projects.
to me, one of the most exciting developments would be revolutionizing the biohackathon. this is actually something i'm working on with a few collaborators and i'll post about it somewhere down the line.
the current biohackathon model, where participants are handed an omics dataset and told to do something "interesting" with it, is quite boring. it produces neat visualizations, but it doesn't produce proofs-of-concepts, MVPs, or founders.
a real biohackathon for the new era would look quite different. this era of biohackathons would leverage the benefits of software, cloud labs, and community labs to challenge competitors to go from problem to prototype. not a theoretical prototype, either - one with preliminary data or a proof of concept. the same way a winning hackathon app can become a YC application, a winning biohackathon project should have a credible path to an accelerator or a pre-seed round.
by making biohackathons actually competitive, with real stakes and real output, we could give the next generation of bio founders the same on-ramp their software peers have had for decades.
of course, none of this is easy, but none of it requires starting from scratch. if implemented in a few communities to start, the movement to democratize our ecosystem could spread like wildfire. in fact, i believe it's already starting to catch on outside of universities - i personally was catalyzed by both my experience in iGEM and by members of our community like elliot roth from biopunk and ulisses santamaria of DMV petri dish.
however, the true litmus test still remains: can we spark biotech's dorm room moment? i think we can.
until next time,
nb.
footnote: i've written an addendum that goes more in-depth about my thoughts on why SaaS going under is inevitable, and what that could mean for science. you can find that here.