Aster operates as an autonomous research lab built around large fleets of AI research agents that work in parallel on open-ended scientific and machine-learning problems. Rather than a single agent iterating serially, the system coordinates many agents to explore, test, and refine hypotheses simultaneously, targeting research tasks that reward broad parallel search.

The lab has demonstrated its approach on public machine-learning and computational-biology benchmarks, reporting results that reach competitive marks far faster and at lower cost than single-agent baselines. Its ambition is to automate the open-ended research loop itself across domains rather than to ship a single vertical product.