micro1

micro1 describes itself as a research lab and data company, but the operating reality is more specific: it produces the human data, environments, and evaluation infrastructure that frontier AI labs use to train and test next-generation models. The threat model here isn't malware, it's model failure - the gap between a system that demos well and one that holds up under adversarial, ambiguous, or out-of-distribution conditions. That makes evaluation rigor and data provenance the load-bearing disciplines, and it's where security-adjacent thinking actually applies.

The technical work spans reinforcement learning environments, AI evaluation and benchmarking, agent evaluation, and robotics data for embodied systems. Three products anchor it: Realm, a reinforcement learning environment platform; Cortex, an AI evaluation and improvement platform; and Robotics, real-world data collection for embodied AI. The company also runs an AI-powered certification process to validate the skills of its expert network, which spans engineering, finance, healthcare, legal, linguistics, and cinematography - the point being domain-qualified human judgment, not generic annotation.

micro1 reports revenue growth from roughly $4 million to over $200 million ARR within a year, and states that its work is done in partnership with frontier AI labs. Operations run out of the San Francisco Bay Area alongside flexible, remote AI training work for domain experts globally. Founding year, team size, and funding details are not publicly specified.