Three problems worth solving, each bigger than the last.
These are the markets we care about, written as the problems to be solved. We are deliberately not prescribing the solutions here. Working out the solutions is what the research is for, and what the fellows are for.
Drugs for the brain are developed without a map of the brain.
Neurology drugs have one of the lowest approval rates in the industry: 5.9 percent of programs that enter Phase 1 reach approval, against 7.9 percent for drugs overall (BIO, 2011 to 2020). Most brain drugs act on a protein on neurons, a receptor, a transporter or an ion channel. Nobody knows, neuron by neuron, where the neurons that carry that protein send their signals. So what a drug will do in the brain is still learned by giving it to animals and then to people.
The GLP-1 drugs, which earned about $70 billion in 2025 (Lilly and Novo Nordisk filings), act on one receptor, or two in the newest of them, and nobody knows, synapse by synapse, which circuits produce each of their effects: less appetite, nausea, less heavy drinking, better sleep apnea. Nausea limits the dose. The industry is choosing the next generation of these drugs by testing receptor combinations in trials of a year or more on thousands of people.
How do you tell, before a trial, what a compound does in a brain, circuit by circuit?
Sources: BIO, clinical development success rates 2011 to 2020; Lilly 10-K, 2025; Novo Nordisk, Q4 2025.
Brain and nervous-system disorders are the leading cause of ill health in the world, and the drugs do not match the need.
3.4 billion people live with a brain or nervous-system disorder (GBD 2021; the count includes migraine and tension-type headache). Infections became treatable once we understood the organisms that cause them and could design drugs against a mechanism. Brain diseases are not treatable that way yet, because the mechanism, which neurons talk to which and what a drug changes in that conversation, has never been seen whole.
The money is already moving. The incretin drugs, GLP-1s among them, are forecast at $200 billion a year by 2030 (J.P. Morgan Research, February 2026). If brain diseases ever become as treatable as infections, the drugs for them will be the biggest category in medicine.
Which mechanisms, in which circuits, should the next generation of brain drugs act on, and how would you know before you build them?
Sources: GBD 2021, nervous system disorders; J.P. Morgan Research, obesity drugs.
Nobody knows whether a complete map of a brain is enough to understand a mind.
The wiring diagram of a worm with 302 neurons has existed since 1986 and still does not predict its behavior on its own. A fly brain of about 130,000 neurons exists as a map and as a model since 2024, and we have measured how much of its published experiments that model predicts: some, not most. A mouse brain has about 70 million neurons and no complete map. A human brain has about 86 billion.
A model good enough to predict what a drug does is a first step toward a model of a brain that runs on its own. Whether a wiring diagram, plus what can be imaged alongside it, is enough to get there is the deepest open question in the field, and the one we most want to answer. The odds of success for this long-term vision are low and the journey will be long. But we are pursuing it anyway, because at a minimum it will further our understanding of the brain, and if we are successful it will be a tremendous benefit to humanity.
Is the brain legible, neuron by neuron, the way artificial neural networks turned out to be, and what would it take to find out?
Sources: White et al., 1986, the C. elegans wiring diagram; Dorkenwald et al., Nature 2024, the FlyWire connectome; Herculano-Houzel et al., PNAS 2006, neuron counts.
Want to work on one of these?
That is what the fellows are for. Pick a problem, bring your discipline, and tell us where you think we are wrong.
Become a fellow