Astro Teller on moonshots: attack the hardest assumption first, kill zombie projects
PeterDiamandis · x · 2026-09-28
Peter Diamandis interviews Astro Teller (Google X) on how moonshots are actually built:
- Start with a huge problem, not a technology; pair the audacity to pursue the unlikely with the humility to assume you're probably wrong.
- Attack the hardest assumption first — what he calls "training the monkey" before building the pedestal.
- The worst outcome isn't failure, but letting a zombie project consume years of time and capital.
- AI is lowering the cost of reaching a moonshot and shortening the distance between a crazy idea and evidence it might work.
More from AGI Musings
- David Hoang: four software types ahead, personal software complements teamwork — davidhoang · 2026-09-28
- Hamel Husain: hiring human agents with AI skills is unreasonable alpha right now — HamelHusain · 2026-09-28
- Burkov: LLMs are imitation engines without human understanding, argues ML author — AlexTensor · 2026-09-28
- ML book author Burkov: anyone claiming 10% extinction odds is nuts or lying — burkov · 2026-09-28
- Reddit Thread Shares Video on What AI Will Do to Jobs and Wages — wilsoniumite · 2026-09-28
- Robotics pioneers split on humanoid timelines: action data, not LLMs, is the bottleneck — Scobleizer · 2026-09-28