Unitree: the "self-evolving" AI that trains its own humanoids
Where it stands: Unitree is developing a "self-evolving" feedback AI intended to cut the need for manual programming so its humanoid robots can improve themselves; no system name, mechanism, demo or deployment timeline exists yet. Founder Wang Xingxing says the goal is robots completing 80% of spoken tasks by 2028, with the self-improving AI named as the route. Watch for an independent demonstration that a robot improves without being reprogrammed, and for measured progress against that 80% target.
Latest
- 2026-10-23 (~12:45 AM ET, MSN syndication): Unitree Robotics is developing a "self-evolving" feedback AI system intended to reduce the need for manual programming and turn AI-model advances into real-world robot capability (msn.com/en-xl). First software-layer program beside the capital lane (the reported $6.2B IPO pursuit, October 14) and the hardware showcase (the H2-X "guardian" promotional video, October 22); no system name, mechanism detail, demo, or deployment timeline in the reporting, so announced-not-shipped applies. Watch: the feedback mechanism's specifics, an independent demonstration of improvement without re-programming, which Unitree hardware consumes it.
- 2026-10-23 (~12:48 PM ET, The Next Web): the self-evolving program acquires a founder-attached, measurable target — Unitree founder Wang Xingxing now expects humanoid robots to complete 80% of spoken tasks by 2028, with the self-improving AI named as the route (thenextweb.com/news). Still announced, not shipped: no demonstration of a robot improving without re-programming in the reporting. Watch unchanged, plus measured progress against the 80% metric.