RL Environments
Proving grounds scored by what actually happened.
Neo Norma
Machines have read everything.
Nature was never asked.
A new standard for intelligence.
Every textbook, memorized.
Every exam, passed.
Then the instrument drifts, the cell divides, the patient does not read the literature.
Most of what science knows was never published. It lives in notebooks, freezers, and archives no model has seen.
What we make
RL Evals
Hard scientific tasks with answers nature already gave.
SFT Data
Expert reasoning traces from working scientists.
Rubrics + Graders
Every answer checked, not guessed.
Agent Trajectories
Long horizon work across real tools and records.
Private Data Partnerships
Records that never leave home, put to work.
What does a model know that was never written down? What happens when the answer comes from the world instead of a page?
We find the record. It never leaves home. Reality does the grading.
Neo Norma. The new standard.
hi@neonorma.si