Observed arrival · 2026-09-19
AutoScientia Wants to Put Scientific Experimentation on Autopilot
AutoScientia presents an autonomous AI system that proposes, runs, evaluates, and iterates experiments across machine learning, science, and engineering.
- For
- AI researchers and experimental engineering teams
- Worth noticing
- The proposed loop uses sandboxed, budget-capped, append-only runs and explicitly addresses flukes through frozen baselines and repeated experiments.
Field notes
The proposed system treats experimentation as a repeatable control loop rather than a one-off assistant: a model writes the next trial, infrastructure schedules parallel runs, and evaluation feeds the result back into the proposer. The page stresses safeguards including sandboxing, budget caps, append-only logs, frozen baselines, repeated runs, and failure taxonomies. Its 1,286-experiment display is labeled an example, not a reported live measurement.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue