Twenty US and twenty-eight Chinese profiles, and four things a robot fleet can share today: weights, intent, ops data and language. The fifth square — one site memory a mixed team reads and writes — is empty, and not because the problem is hard.
The best-documented robot releases in the field publish success rates from 32 to 92 percent with no trial count behind them, and a 0.54-million-parameter policy nearly matches a 4.1-billion-parameter one on the benchmark everyone quotes. At the trial counts this field actually runs, a published success rate cannot separate a real improvement from noise. Here is what to ask a robotics vendor for instead, and the ten rules we have bound ourselves to before we publish a number of our own.
Singapore's cleaning sector is not short of labour. It is capped: a 35 percent Dependency Ratio Ceiling, an S$800 monthly levy at the top tier, and a wage floor that is a condition of the operator's licence. Meanwhile the two statistics the industry quotes trace back to a fundraising announcement, a manufacturer's marketing page and a vendor whitepaper, and nobody in either chain can name a measurement.
ISO 10218-1:2025 excludes medical robots, healthcare robots and service robots the public can access, and names no standard that would reach them. ISO 3691-4:2023 carves out the additional hazards of trucks intended to operate in public zones; ISO/CD 25785-1 defines its industrial environment as a workplace where the public is excluded or restricted from access; both name ISO 13482 as the standard for what they exclude. Three standards push a legged robot in a hospital out, two of them pointing at a fourth whose in-force edition covers three personal care robot types it is not one of, while the edition written for service robots is an unpublished draft. That is why a safety file for that machine is built from ISO 12100 upward, not claimed.