Computing Life · YageMay 12, 2026, 07:00 UTC+8

How AI Caused and Fixed My Insomnia

The author used AI to build a HealthKit export tool and run multivariate regression, found that post-dinner AI use correlated negatively with sleep duration, and added 1 hour 40 minutes of average nightly sleep after avoiding AI for several weeks.

Why it matters: HKR-H/K/R all pass: the personal reversal is clickable, the HealthKit/regression setup adds testable detail, and sleep loss hits AI practitioners directly. Scope is anecdotal, so it stays at the featured floor.

Read the original ↗Export Markdown

Take

The sharp part is not that AI diagnosed insomnia; it is that AI turned “light computer use after dinner” into sustained cognitive labor. The author spent five minutes generating a HealthKit export app, pulled Apple Watch, iPhone, Withings-style health streams, then ran multivariate regression. The strongest negative correlate was the last evening AI session; after several weeks off post-dinner AI, average sleep rose by 1 hour 40 minutes.

I buy the mechanism more than the causal claim. This is one person, a few weeks, and the article does not give p-values, coefficients, or a randomized crossover. Still, every heavy agent user knows the loop: run several sessions, poll outputs, split tasks again, keep reading. There is no idle wait. Compared with YouTube or games, late-night AI looks less like screen time and more like sneaking a strategy meeting into bed.

Source: Computing Life · Yage · yage.ai
Published: May 12, 2026, 07:00 UTC+8 · May 12, 2026