Sharpa Co-Founder Li Yifan: General-Purpose Humanoid Robots Are Either All-Round or Useless

Sharpa co-founder Li Yifan argues that general-purpose humanoid robots must achieve full task autonomy without modifying environments, warning against the 'phone booth' trap of specialized automation where robots only handle isolated steps. Citing a recent DQ ice cream trial with 55 steps completed on existing equipment, he asserts that partial success (90%) is a dead end because remaining human intervention negates value and exposes the system to specialized competition. The piece outlines Sharpa's roadmap for scaling from single-scenario mastery to cross-task skill transfer, distinguishing between data-driven generalization and mere repetition of simple tasks.