Halt

Halt learned to plan for failure the day a robot that couldn't fail safely scared everyone.

He was young, and he'd built a fast little robot with no plan for trouble at all. When its steering broke mid-run, it didn't stop — it just kept going, full speed, blind, careening around the room until it crashed into a shelf. Nobody was hurt, but everyone had scrambled out of its way, hearts pounding.

His grandmother, an old armadillo with a scarred, sturdy shell, sat with him.

"The steering broke," young Halt said, shaken. "That wasn't my fault!"

"The breaking wasn't your fault, little one," his grandmother said gently. "Things break; that's the world. But what it did when it broke — kept racing, blind and dangerous — that was a choice you didn't make. You never told it what to do when things went wrong, so it did the worst possible thing: nothing but keep going." She tapped his shell. "A good builder assumes everything will fail, and plans the failure gently. If the steering breaks — stop. If the battery's low — return, or park safely. If a sensor goes dark — slow down, don't guess. If anything's wrong at all — fail toward safe, never toward danger." She smiled. "And here's the beautiful part: once your robot always fails safely, you can be braver. You can try wild, fast, ambitious things — because you know that if it breaks, it'll stop gently, not hurt anyone. Fail-safe design isn't gloomy. It's what makes bold building possible."

Young Halt never built a robot without a failure plan again. He knew: you can't stop things from breaking. But you can always decide how they break.

He walked to RoboForge at twelve, because a place that built robots ought to build them to break safely.

Servo, his mentor, met him at the workbench. "What does your robot do when it breaks?"

Halt didn't answer with a speech. He took a robot and, mid-task, yanked a wire loose on purpose. The robot — which he'd designed fail-safe — instantly noticed the fault, gently lowered what it was holding, powered down its motors, and blinked a calm red light: I have a problem, I've stopped safely, please check me. "There," Halt said. "It broke. And instead of grinding or racing or dropping something, it stopped, safely, and asked for help. That's the behavior I care about most. Not what it does when it works. What it does when it doesn't."

Servo watched the robot fail gently. "You are appointed," he said.

At Halt's workbench, he taught fail-safe design by breaking things on purpose. He'd have the kids build a robot, then he'd sabotage it — jam a wheel, unplug a sensor, drain the battery — and ask: "What did it do? Was that safe?" A robot that kept forcing a jammed wheel (overheating) — not safe; add a rule: "if the wheel won't turn, stop." A robot that guessed blindly when its sensor died (crashing) — not safe; add a rule: "if you can't see, slow down and stop." Then he gave the working rules: assume every part will fail eventually, and decide in advance what the robot does when each one does; the golden rule is fail toward safe — when something's wrong, stop, slow, park, or power-down gently, never grind, force, race, or guess; and a clear "I have a problem" signal (a light, a beep) so a human can help beats a robot that hides its failure and struggles on. "A student thought planning for failure was 'giving up before you start,'" Halt said, curling calm. "But it's the opposite. Because his robot always failed safely, he could finally try the bold, fast, risky things he'd been afraid of — knowing that if they broke, they'd break gently. Fail-safe design doesn't make you timid. It makes you free to be brave."