Who audits the experiment when the experimenter is the model itself?
Opportunity
Autonomous AI systems are being deployed across biology, materials science, and software engineering to design experiments, run them, and write findings with minimal human review per step. A 2026 survey of AI scientist systems found that only 38 percent release the seeds or execution traces needed for a reviewer to reproduce a run, and exactly one of nine fully autonomous systems has been externally validated. There is no mechanism to verify that the steps a model reports match what was actually executed, or that its statistical summaries correctly reflect the underlying data. Human spot-checking does not scale to the volume these systems now produce.
Why it matters
Scientific findings compound through citation, so one convincing but unverifiable AI result can corrupt downstream research faster than any correction process can catch.
How I score the opportunity
The Opportunity Score is my own read, not a measurement: how much it hurts, how often it bites, and how little exists to solve it today. Higher means I think it is more worth building.
How much pain it causes when it shows up.
How often people actually run into it.
How little good tooling exists for it today.
More problems worth solving
Why does every AI app forget me the moment I close the tab?
AIWhy is learning a new field still gated by knowing what to ask?
AIWhy can a non-expert not verify what an AI just told them?
AIWhy do we test models on benchmarks but ship them on vibes?
AIWhy do AI agents have no memory of their own mistakes?
AIWhy can't I audit what a model was actually trained on?