We are an applied AI lab.
Most of the AI being built today is being built to win a market. We are interested in the AI that gets built when no one is winning. The triage assistant for the free clinic. The reading layer for a small newsroom. The planner for a community land trust thirty years out.
These are tools. They will not become products. They will sit upstream of someone's job and quietly make the job possible. That is the work.
Three things we believe.
We believe intelligence belongs in infrastructure, not in another product. A model embedded in someone's workflow, where they don't have to think about it, is doing more good than a chatbot they have to remember to open.
We believe the lab format matters. A lab is allowed to try things, publish what didn't work, and pause a project that isn't ready. A startup isn't. We chose to be the former.
We believe the work should outlast the people doing it. That's why this lab is anonymous on purpose — and why every project we ship comes with a write-up in case someone else wants to take it further.
What we will not do.
We will not build surveillance tools, persuasion systems, autonomous weapons, or anything whose primary mechanism is taking choice away from a human. We will not work for clients whose incentives we cannot inspect. We will not race to ship something we are not yet sure is doing more good than harm.
We will sometimes be wrong about whether a project belongs in this lab. The remedy for that is to write down our reasoning when we start, and write down what changed when we stop.
How to work with us.
If you want to contribute, look at the experiments. The page for each project tells you what kind of help we're looking for. Send a note. We read everything.
If you want to fund this work, the same goes. We are deliberately small and intend to stay that way; what we want from funders is patience and a shared definition of "worth doing."
If you are working on something adjacent to what we do, we want to hear from you even if there's no overlap to act on yet. The lab is deliberately small but not deliberately closed.