Our vision

Make scientific progress
scalable with compute.

We’re building AI scientists that make discoveries and improve the way discovery happens.

Scientific progress depends on how many ideas we can explore, how well we test them, and how much we learn from each experiment. Gnosys is building research teams that expand that capacity with compute.

The vision is a compounding cycle of discovery: AI scientists develop better models, tools, and learning methods, then use those advances to build the next generation of scientists. More research capacity means more progress across AI, medicine, materials, energy, and beyond.

Teach discovery through discovery

The next frontier is training environments built around known science. Start a researcher with knowledge from before a discovery. Let it propose experiments, observe results from recorded data or validated simulations, and work toward the discovery itself.

These environments give researchers practice forming hypotheses, choosing experiments, and learning from evidence. We evaluate those skills on unfamiliar problems, keeping the answers outside the researcher’s starting knowledge.

Improve the researcher, then its successor

A discovery solves a problem. A better scientist expands what we can discover. Our research targets the skills behind that progress: choosing informative experiments, building better measurements, and directing effort toward the most valuable questions.

Better scientists → better tools and learning methods → better scientists.

Our next experiments put an improved researcher in charge of developing its successor. We measure progress across generations on fresh problems, tracking both research quality and cost.

Start with AI. Build toward science.

Our first environment is AI and ML, where many experiments run in software. We’re building coordinated research agents and methods for measuring progress when feedback is scarce or imperfect.

Fraud detection brings this work to a concrete business problem. Teams bring their data and domain judgment; Gnosys develops models they own. The research continues as their products, data, and adversaries evolve.

We’re building the foundation for a broader research engine: reusable methods, stronger measurements, and environments where AI scientists learn through experimentation.

The research ahead

Three priorities drive the next stage:

  1. Better feedback. Develop researchers that build and refine their own measurements, and evaluate the gains against the cost.
  2. Methods that transfer. Train research skills that carry across tasks and evaluate them on unfamiliar problems.
  3. Researchers that improve their successors. Build and test the handoff between generations.

Our roadmap extends from AI and ML into historical scientific environments, then into new science and physical experimentation. Each stage expands the range of questions our researchers can tackle.

Join us in building this next generation of AI scientists. We’re bringing together investors, researchers, and collaborators who share the ambition to accelerate discovery.

We’d like to use one cookie to count unique visits. Privacy