Sara Hooker's Adaption Labs raises $50M to build AI that learns continuously without retraining
Feb 9, 2026 with Sara Hooker
Key Points
- Adaption Labs closed a $50M seed round three months after founding to build AI systems that continuously adapt to incoming data without costly retraining cycles.
- Co-founder Sara Hooker is deliberately rejecting the frontier-model paradigm that dominated her prior research career, betting instead that static models requiring prompt engineering are a dead end.
- Hooker measures progress by eliminating prompt tuning rather than benchmark scores, treating continuous real-time learning as the mechanism to close the gap between model capability and actual usability.
Summary
Sara Hooker founded Adaption Labs three months ago with cofounder Sadeep and immediately closed a $50M seed round. The company rejects the frontier-model paradigm that defined Hooker's earlier research career. Instead of chasing scale, Adaption Labs is solving how to make AI systems that continuously adapt to incoming data without costly retraining cycles.
Hooker frames the problem as a gap between model capability and actual usability. She measures progress by eliminating prompt engineering, which she views as a symptom of systems that fail in real use. Continuous learning that happens in real time, without retraining, closes that gap. Static models that require prompt tuning are a dead end. Systems that evolve with user data are the viable product direction.
The $50M raise in three months reflects investor conviction in both the problem and the team's execution capability on continual learning at scale. The segment does not disclose lead investors, fund sizes, or valuation.