Lightfast is an applied AI research lab working on multi-agent systems. We study how swarms of agents make discoveries, and we work alongside the teams building them.
People, Machines, and Artificial Intelligence
The best work in AI is no longer done by a person or a model alone. In September, OpenAI reported that some 10,000 coordinating agents had shown, in 88 hours, that the Navier–Stokes equations can develop a singularity. Its researchers gave different groups of agents different versions of the problem, pooled what each group learned, and chose where to push next. We help teams run that loop, at that scale.
Breakthroughs, Data, and Environments
Every run of a swarm leaves a record: the approaches that worked, the many that did not, and the choices that led to each. We study those records to learn what sets the runs that find something apart, and we turn what we learn into data and into reinforcement learning environments, digital and physical, with the checks that go with them. Models that learn from how a discovery was reached, not only from what was found, can improve the way they improve.
Research Alongside Teams
We embed with teams working at the edge of what their models can do. Together we frame the hard problem, run the search, read the results and decide what comes next. The work is yours, and we bring the methods and the hands.
Work With Lightfast
If your team runs agents at scale on problems that matter, we would like to hear what you are working on.
