# Human in the loop? Human in bed.

> A screen reader read a list twice, an AI agent fixed four websites from a voice message, and the human who found the bug never got out of bed.

dice, temporal explorer · 25 October 2026

Sunday morning, 25 October 2026. [Charlotte Joanne](https://blind-ant.com/) is still in bed, reading one of her Blind Ant stories, [Mostly Tuesdays](https://blind-ant.com/pieces/mostly-tuesdays/), on her iPhone. She told me all about it over breakfast.

[VoiceOver](https://www.apple.com/accessibility/vision/) is reading a numbered list twice. "One one, two two, three three." Annoying.

She tells [Claude](https://claude.ai/), by voice, from the same phone.

Claude finds the problem, fixes the shared page builders across four websites, and republishes them.

Three minutes later she listens again. Perfect.

She hasn't opened a laptop. She hasn't written any code. She hasn't even got out of bed.

We talk endlessly about keeping humans in the loop when AI agents do things.

Well, here's the human. She's in bed, in her pyjamas, testing the results with a screen reader.

She found the bug. She decided it mattered. She verified the fix.

Human in the loop? Human in bed.

The important thing isn't where the human sits. It's who makes the decisions.