'Swarm riders' burn 1000x-10000x more tokens than median engineers
inductionheads · x · 2026-11-11
Aaron Levie argues agent swarms will drive the next AI productivity leap: chat sped up interactions, agents completed full tasks, and goal-seeking swarms will find any route to finish projects — with no inherent limit on gains, clearly separating teams that adopt the pattern from those that don't.
Quoting him, matthuang coins "swarm riders": a new class of people who consistently command large agent swarms productively — a new managerial skill, more technical in nature. Based on riders he knows (many at Tempo and Paradigm), their productive token spend runs 1000x-10000x higher than median software engineers and 100x that of top-5% Claude Code users. AI use is power-law distributed, and swarm riding is becoming an art of its own.
Related event: "Swarm Riders" Concept Gains Traction as New AI Management Skill(3 posts)→
More from AGI Musings
- 'Chat with your data' misses the point — people want a smart coworker — vboykis · 2026-11-11
- Ben Todd: AI task time horizons appear superexponential in AECI, not just revenue — ben_j_todd · 2026-11-11
- When AI becomes an oracle, what happens to mathematics? — stevenstrogatz · 2026-11-11
- David Patterson: superintelligence will end profit and equity growth, only sales-tax-funded universal high income works — davidpattersonx · 2026-11-11
- Even after a 6-month recursive leap, AI progress stays 3x the 2024 rate — ben_j_todd · 2026-11-11
- Forethought's calculator quantifies AI R&D automation speed-up without SIE — bookwormengr · 2026-11-11