Solo Developer's Robot Safety Code Halts Machines in Under a Millisecond

A small, free piece of software designed to be a robot's 'reflex layer' just got put through independent testing — and its headline promise survived. The software, called QERRA-HSR, sits between a robot's decision-making brain and its motors. When danger appears, it is supposed to slam the brakes immediately. An outside team measured the delay between detecting danger and sending the stop command: 0.0 milliseconds. That is not 'fast' — that is effectively instant.

The testers also found genuine problems, and the developer fixed them the same day. First, a cooldown timer bug meant that if sensor readings changed while the robot was calming down, the clock never restarted — so it could wrongly announce 'all clear.' Now it demands a full, unbroken second of calm. Second, a setting called the task-interruption flag was accepted by the code but never actually used; now routine jobs restart on their own after danger passes, while delicate jobs freeze until a human confirms it is safe. Third, the safety timer was chained to the real computer clock, which made testing awkward — it now accepts a simulated clock. Fourth, and most honestly, the console printed 'STOP' the instant danger hit, but momentum still carries a robot about 4.5 centimetres before it fully rests. The stop command really is instant; physics is not. The documentation now says so plainly.

Why should you care? Because robots are moving from factory cages into warehouses, hospitals, farms and eventually homes, and the hardest problem is not intelligence — it is trust. A robot that decides well but stops slowly is dangerous. This project's selling point is that its rules are 'deterministic' and auditable: they behave predictably and a human can read exactly why a decision was made, rather than trusting a black box.

The catch is significant. This is one self-taught developer working solo, using AI tools as coding and debugging assistants. All 20 automated tests pass, and the live code is updated — but the proof comes from simulation, not a physical robot in a room with people. Promising engineering, still early days.