We’re not trying to throw our weight around here… but we’ve been in the business of education and training for more than 75 years. NIST held its first Weights and Measures School in 1948. Classes were focused on practical training, including work on actual weighing and measuring devices (as seen in image 1/2). (Side note: While our first class was in 1948, our weights and measures work can be traced back to 1836!) Today, NIST’s Office of Weights and Measures continues to offer in-person and online trainings, educating weights and measures inspectors from all 50 states and U.S. territories on key tasks that range from ensuring you get what you pay for at the pump to confirming that grocery scales are accurate. In-person classes take place in our Weights and Measures Lab (as seen in image 2/2).
This kind of training is so essential, awesome to see NIST at it after all these years! 👏
There may be a remarkably direct line from this history to the next measurement problem. Weights and measures made physical commerce trustworthy by replacing “trust the claim” with standards, traceability and reproducible measurement. AI agents are beginning to create a similar problem at the execution layer: what exactly was requested, what rule governed it, what state changed, and can an independent party reproduce the result? I’ve been working on a deterministic execution architecture that separates the probabilistic planner from commit authority. The planner may propose an action; a deterministic work engine evaluates fixed/versioned constraints and resolves OPEN or HALT, producing a cryptographically bound execution receipt and replay path. In that sense, the problem starts looking less like another AI benchmark and more like metrology for machine execution: defined conditions, observable transitions, traceability, conformance and reproducibility. NIST’s weights-and-measures history seems like an unusually appropriate place to recognize that connection. The full workstation is here: public-contract.github.io