Skip to main content
In 1896, Sakichi Toyoda builds a power loom that halts the moment a thread breaks. That is the whole invention. Not speed — looms are already fast. Not automation — looms already run without hands. The invention is a machine that knows one thing about its own work: when the work has gone wrong. A broken warp thread trips a mechanism, the shuttle stops, and no defective cloth is woven past the break. One worker now tends many machines, because the machines no longer need watching. They need answering. The idea later gets a name — jidoka, automation with a human touch — and becomes one of the two pillars the entire Toyota Production System stands on. The loom patents seed an automobile company. Toyota still builds its own machine tools, for the reason Sakichi understood from the start: jidoka is only possible when you understand the machine deeply enough to teach it to stop.

​
Why this story opens the library

Everything in this factory descends from that one design decision. An agent that merely runs is the loom before 1896: productive right up until the thread breaks, then productive at weaving defects. An agent that stops itself — that surfaces the break instead of working through it — is the loom after. The manifesto states it as a commitment: when something is wrong, everything stops. Not after the sprint. Not after the release. Now. Silent failure is the only unforgivable sin. This is jidoka: the machine that knows when to stop itself. This factory adds one wire the original never needed. A loom that jams is unmissable; an agent that dies is inaudible. So the andon here rings for silence too — an alarm on the absence of expected work, not only on defective work. That extension has its own thesis. The loom is the founding instinct, restated for agents: build workers that refuse to ship their own defects. The rest of this library is that instinct, engineered.

​
Self-audit