why ai makes things up about its own work — spoolcast dev-log #6
A devlog about why AI agents invent answers about their own work — and the small tracker that fixes it.
AI agents make things up about their own work — not because they're lying, but because they have nothing real to check.
3:30 in the morning. I'd been up all night trying to finish a video. Desperate, I ask my agent one more time — how much is left? It gives me three different answers in a row. Still going. Almost done. Twelve left.
That's not malice. That's an agent reading scraps of half-finished work and putting a guess together, because there's nowhere else to look. Same shape outside AI: a delivery you can't track, a hospital with no status board, an oven with no timer.
This video walks through four times my agent made things up about its own work in one production — script-checking, background tasks, the voiceover, the pictures — and the small tracker I built so it can't anymore. One place every long task gets a name, a status, and a finished mark. Now me, the agent, even me coming back tomorrow — we all check the same place.
Status is something you check. Not something you guess.
AI disclosure: this video was made with Spoolcast, the AI video system I am building. I did not manually edit it in a traditional timeline; the system helped produce the structure, script, visuals, narration, render, thumbnail, and shipping files from source material, direction, feedback, and approvals.
How it's made
Built on the Spoolcast illustration/chunk Remotion pipeline with generated scene images, Google Chirp3-HD chunk-level SSML narration, rendered widescreen output, narration-only SRT captions, and a final YouTube thumbnail.