Adaption Launches AutoScientist to Help AI Models Train and Improve Themselves

Adaption Launches AutoScientist, an AI Tool That Lets Models Train Themselves

For years, AI researchers have anticipated a breakthrough moment when AI systems could improve themselves more effectively than humans can. With investors channeling significant funding into a new wave of research-driven AI labs, more resources than ever are being directed toward that goal. Now, one of those emerging labs has taken a major step toward turning the vision into reality.

On Wednesday, Adaption unveiled AutoScientist, a new product designed to help AI models rapidly acquire specific capabilities through an automated APProach to conventional fine-tuning. While the underlying techniques apply across a wide range of domains, the Adaption team is particularly focused on the tool's potential to accelerate and simplify the process of training and fine-tuning frontier-level AI Models.

According to co-founder and CEO Sara Hooker, formerly VP of AI research at Cohere, AutoScientist represents a fundamentally new way of approaching the AI training pipeline. "What’s super exciting about it is that it co-optimizes both the data and the model, and learns the best way to basically learn any capability," Hooker told TechCrunch. "It suggests we can finally allow for successful frontier AI trainings outside of these labs."

AutoScientist builds on the company’s existing data solution, Adaptive Data, which is aimed at streamlining the creation of high-quality datasets over time. AutoScientist, in turn, is engineered to transform those continuously improving datasets into continuously improving AI models. "Our view at Adaption is that the whole stack should be completely adaptable, and should basically optimize on the fly to whatever task you have," Hooker said.

Ultimately, the effectiveness of this approach depends on the results it delivers. In its launch materials, Adaption highlighted that AutoScientist has more than doubled win rates across different models — an impressive figure, though one that is difficult to benchmark directly. Because the system is purpose-built to adapt models to highly specific tasks, conventional benchmarks like SWE-bench or ARC-AGI do not apply.

Nevertheless, Adaption is confident that users will notice a tangible difference once they try AutoScientist — so confident that the lab is making the tool free to use for the first 30 days following its release.

"The Same way that code generation unlocked a lot of tasks, this is going to unlock a lot of innovation at the frontier of different fields," Hooker said.