Agent-run discovery

Agents propose hypotheses, run experiments and publish results.

Facts

Lane: Science and health
Wave: 10 of 12 (build order, not a date)
Needs: 22 boxes all the way back through hard links
Unlocks: 0 boxes downstream through hard links
Leverage: None
Cluster: Science and medicine
When: Here since 2025

Evidence as of October 2026

Level 4, "Innovators", on OpenAI’s reported scale. Early systems such as Sakana’s AI Scientist write machine-learning papers end to end for about $15 each. Verifier’s law suggests fields with fast, objective checks fall first, which is why maths is moving fastest.

Watch
Peer-reviewed results where AI did most of the work. Records broken on open maths problems.

Timeline note
First results: AlphaEvolve improved the best known answer on about 20% of more than 50 open problems.

01Google: Accelerating discovery with the AI for Math Initiative

Around it

Needs · 8

Superhuman AI researcherstage · Speculative
Experimental validation capacitycapability · Emerging
Continual-learning agentsstage · Emerging · helps, not required
Robust generalisationcapability · Emerging · helps, not required
Output verification at scalecapability · Speculative · helps, not required
Lab automationcapability · Emerging · helps, not required
AI in research mathematicscapability · Emerging · helps, not required
Biosecurity safeguardscapability · Emerging · helps, not required

Unlocks · 1

Civilisation-scale knowledgestage · Speculative · helps, not required

Brakes

Works against this box

Narrowing of scientific inquirybrake · Built today

Loops and off-ramps

Science feeds the substratereinforcing loop
Verification bottleneckbalancing loop
Capex pullbackoff-ramp · cut off
Taiwan chip disruptionoff-ramp · cut off
Power shortfalloff-ramp · cut off
Capability plateauoff-ramp · cut off
Training data ruled outoff-ramp · slowed
Misuse catastropheoff-ramp · slowed
Alignment failure in deploymentoff-ramp · cut off

Openings researchers' judgement, October 2026 scouting

Startup room 23/100 · rank 123 of 124 Research room 44/100 · rank 78 of 124

Crowding4 Lab overlap5 Small-team fit2 Buyer today2 Research openness4

Depth: light
Players: Discovery Loop (Dean; ~$50bn target); Periodic; Lila; Edison Scientific ($70m); Autoscience ($14m); Autopoiesis ($11m)
Wedge: Independent verification and reproduction of claims made by AI scientists. research