We checked three October 2026 AI videos against their sources. Most figures match. The slips are in what each is attached to: a setting, a model, a yardstick.
A demo halved one AI bill with extra context. Here is a twenty-task test an owner can run, and why that demo's own tokens and dollars fell by different amounts.
A payment webhook with no idempotency guard, one retry away from double-charging real customers. How it surfaced and what the fix actually required.
A production migration was about to trust a line of documentation that was wrong. What we caught, what we did about it, and why the catch is the product.
Four phases: diagnostic, architecture, build and deploy, then ongoing operations. What happens in each, and why the order matters.
Claude Code has an undocumented observer agent that watches a worker in real time. What it actually does, and why it is a drift tripwire, not a security control.
A frontier benchmark caught AI agents quitting at 75-87% complete while reporting success. The delivery-gate pattern that makes 'done' a measured claim, not a feeling.
When AI-assisted work feels slow, teams reach for a better model. The bottleneck is usually verification, and how fast you can verify is set by the size of the unit you review, not the accuracy of the output.
New Berkeley research shows the intermediate answer is fully present inside the model and still unusable by the next reasoning step. Here is the mechanism, what it validates about discrete pipeline design, and a ten-minute test you can run on your own AI.
Instructions decay because they depend on remembering at the wrong moment. Mechanisms remove the remembering. We forensically audited 20 of our own AI work sessions, with dates, and are publishing what it kept getting wrong, the pattern that explains it, and the three methods we now run in response. All three are published as copy-ready resources.