The Reliability Gap
Cloudadorn Academy
Why does AI that dazzles in a demo stall in real work? Shelley works in operations at Hartwell, a furniture retailer made up for the show, where her boss Gordon wants an AI agent handling customer returns by spring and forwards her every headline and vendor deck on the subject. Rob has read the documents behind those headlines. Across eight episodes they take one claim at a time back to where it came from: the difference between an assistant and an agent, METR's way of measuring how long a task these systems can manage, why a chain of steps that each usually work still goes wrong, three adoption surveys that seem to disagree, Klarna's press release set against its filing, the MIT report behind a famous ninety-five percent, the postmortems that exist and the ones that don't, and what a 1990 paper about electricity in factories says about where this is heading. Two questions run through all of it, and they work in any meeting: at what reliability, and over what length of task? No tech