AI tutoring at scale

Personal tutors for every learner.

Facts

Lane: Interfaces and society
Wave: 5 of 12 (build order, not a date)
Needs: 8 boxes all the way back through hard links
Unlocks: 1 boxes downstream through hard links
Leverage: None
Cluster: Assistants and the public
When: Researcher estimate: 2026 to 2032

Evidence as of October 2026

A World Bank pilot in Nigeria found six weeks of after-school AI tutoring raised scores by about 0.3 standard deviations, roughly 1.5 to 2 years of typical progress, and Harvard physics students learned more with an AI tutor. Other trials show small or no effects, and low usage kills results.

Watch
Effect sizes in large randomised trials, and how much students actually use the tutor.

Timeline note
My estimate. Trials show large but uneven gains, and scale depends on whether students keep using the tutor.

01Five recent AI tutoring studies

Around it

Needs · 4

Conversational assistantsstage · Built today
Real-world task evalscapability · Emerging · helps, not required
Persistent memorycapability · Emerging · helps, not required
Low-price AI for emerging marketsstage · Emerging · helps, not required

Unlocks · 1

Expertise pipelinecapability · Speculative

Loops and off-ramps

Expertise pipelineharmful loop
Capex pullbackoff-ramp · cut off
Taiwan chip disruptionoff-ramp · cut off
Power shortfalloff-ramp · cut off
Training data ruled outoff-ramp · slowed
Misuse catastropheoff-ramp · cut off

Openings researchers' judgement, October 2026 scouting

Startup room 33/100 · rank 100 of 124 Research room 52/100 · rank 57 of 124

Crowding4 Lab overlap4 Small-team fit3 Buyer today3 Research openness5

Depth: light
Players: Khan Academy Khanmigo; frontier-lab education features
Wedge: Engagement research: getting students to use tutors productively, the binding constraint in the NBER RCT. research

01NBER w35620: Khanmigo RCT