Physical AI Lab // Episode 0.5 // One-pager
The humanoid robot boom
Demo, prototype, mass production: what's the difference?
Three key points
- A demo is "the best take": often after many tries, sometimes with a person remote-controlling behind the scenes. It tells you the upper limit, not how many tries out of ten succeed on a normal day.
- Tesla's Q2 2026 update, page 6: first-generation Optimus lines are being installed; initial builds are planned for training data collection and further functionality development. The table lists Construction. A maker's report for that period, not independent proof of current output, delivery at scale or reliable long-term work.
- One industry analysis discusses 95%+ uptime and 30–90 minutes of continuous operation (M). A percentage and duration cannot be directly converted, and this is not a measured benchmark for all models. Check O/R/M sources and label makers' performance numbers as self-reported.
Evidence: O official · R research · M media
| M | Best take / teleoperation | An industry analysis says demos are often best-case results after many attempts, and many "autonomous" demos rely on teleoperation; not a check of every video (M) |
| M | 95%+ / 30–90 minutes | An industry analysis discusses uptime share and continuous runtime separately; not directly convertible, not a measured benchmark for all models (M) |
| O | Construction · Q2 2026 | Tesla's Q2 2026 update, page 6: first-generation Optimus lines are being installed; initial builds are planned for training data collection and further functionality development. The table lists Construction. A maker's report for that period, not independent proof of current output, delivery at scale or reliable long-term work. |
Further reading
- Tesla · Q2 2026 Update · p. 6 (reporting period ends 2026-06-30)
https://assets-ir.tesla.com/tesla-contents/IR/TSLA-Q2-2026-Update.pdf - Challenges in Humanoid Robotics (RoboZaps, reliability and cost)
https://blog.robozaps.com/b/challenges-in-humanoid-robotics - Physical AI Frontier Radar (how evidence is graded O/R/M)
https://noteflowai.github.io/physical-ai-radar/