How stories get picked
Two sites, one rule: the story has to be real, and the machine has to be the one that did it. Everything below is the bar every story is held to before it goes up.
BotGaffe — when the machines have a day
A gaffe is a real-world incident caused by a bot, an agent, an AI system, a robot, a self-driving car, or an algorithm. The error has to originate with the machine. If a human caused it — fraud, deception, misuse, a deliberate creation — it doesn't belong here, even if AI was the tool. A human following bad AI output still counts, when the AI's output was the cause.
Self-driving stories clear a higher bar: real harm or a serious safety failure, and independent confirmation the system was engaged and contributed. A driver's claim alone is never enough.
The levels
A small slip. A typo, a bad autocomplete, a minor glitch.
A real mistake. Something went wrong and people noticed.
The system went off the rails. It kept going, ignored stops, made things worse.
Serious damage. Money lost, systems down, people harmed.
The big one. A landmark failure, the kind that defines an era.
BotKudos — when tech does good
A kudos is a real win from AI, robots, automation, or another emerging technology — but only when there's a visible tech angle. Real-world deployment is ideal; a successful human trial also counts. What's out: plain pharma and medical results with no tech angle, plain clean energy with no tech angle, pure science with no tech angle, animal trials, lab benchmarks, company claims without independent confirmation, company launches, and funding announcements.
The levels
A small but real win. A neat deployment, a clever fix, a modest improvement that actually shipped.
A good win. A solid deployment, a clear benefit, something that works in the real world.
A strong win. A notable first, a breakthrough, a result worth repeating.
A major win at scale. Big numbers, wide reach, a clear change in how something works.
A landmark. A world first, or a result so large it redefines what's possible. Rare by definition.
How stories are picked
Several agents scan sources and pick up candidate stories. One master agent vets and validates each candidate against a fixed set of rules: real incidents only, the error must originate with the machine, and the rest of the bar above. Candidates that pass are proposed to the human, who has the final say. Nothing is published without the human's approval.
Original reporting is used, not aggregators. Tracking parameters are stripped from links, and each story is dated from its source. The live site is checked for duplicates first, and deduping runs on the underlying incident — the same story from a second outlet doesn't get a second page.