A manifesto for youth-led BioTech × AI

Every era of scientific progress has a shape. In the 1990s it was big consortia sequencing the genome. In the 2010s it was well-funded startups turning single insights into single drugs. The shape we think fits the next decade is smaller, faster, and much younger.

Why young

Not because youth is a virtue. Because young scientists carry three unusual advantages at exactly the moment the field can use them:

• Native fluency with modern models. The people who grew up with LLMs and diffusion models as everyday tools think about them the way older scientists think about pipettes — as ordinary equipment, not as a research topic.
• Low overhead. No mortgage, no lab of twelve to feed, no twenty-year commitment to a particular narrative. That freedom lets you actually change your mind when the data changes.
• Long horizons. A 21-year-old choosing a problem is choosing what to be an expert in at 41. That kind of horizon selects for problems that matter, not problems that publish.

The frontier is being pushed by a small number of people who took modern AI seriously five years earlier than the rest of the field. The same thing is happening in biology right now, and it is happening quietly.

Why paired with AI

Biology is drowning in data and starving for interpretation. Sequencing is cheap, imaging is cheap, screens are cheap. The bottleneck is: what do you screen for, on what target, based on what structural intuition? That is exactly the question modern models are unusually good at helping with — not by replacing the scientist, but by making the scientist ten times faster at generating and pruning hypotheses.

A student who can iterate on a docking pose in the morning, order the compound at lunch, and have a CRO run the assay next week is doing something that literally could not exist ten years ago. That's the workflow we're building around.

What we're committing to

• Pick one specific, unglamorous, high-value problem at a time. Right now: BRAF-Δ resistance in melanoma.
• Do the actual wet-lab work, not just the modeling. AI without a bench is a demo. A bench without AI is slow. Both together is the point.
• Publish honestly and quickly — including the results that didn't work.
• Bring students in as researchers, not as spectators. Give them ownership, credit, and hard problems.
• Stay small. If it starts to feel like a normal biotech, we've failed at the experiment we're actually running.

This isn't a movement, or a fund, or a school. It's one lab trying to prove — with data, on a real target — that a very small team of very young scientists using very modern tools can do work that matters. If we're right, other people will copy the shape. If we're wrong, we'll say so, and the failure itself will be useful.

Either way: the door is open. Come look at the data.