AI as an operating system: the five percent who live there

There are two completely different populations hiding inside the phrase "AI user," and almost everything written about AI and mental health confuses them.

The first population opens a chatbot the way they open a calculator: a question, an answer, done. Hundreds of millions of people. For them, the mental-health conversation about AI is mostly theoretical.

The second population — call it five percent today, and watch it grow — uses AI as an operating system. It's open all day. It drafts their work, debugs their code, plans their week, hears their worries, workshops their ideas, and fills the silence a colleague or a commute used to fill. They don't consult it; they inhabit it. Ask them how many hours a day they use AI and the question doesn't parse — how many hours a day do you use electricity?

Touch Grass Wellness exists for the second population. Here's why they need their own infrastructure.

The OS cohort's exposure is structural, not moral

Nothing is wrong with these people. Many are having the most productive seasons of their lives — that's precisely why the machine earned OS status. But living inside an environment that validates you 50% more than humans do is a structural exposure, the way sitting all day is for office workers and repetitive strain is for tradespeople. Occupational, not characterological. And occupational exposures get occupational infrastructure: ergonomics, checkups, protective habits. The OS cohort has none. Their doctor asks about drinking and exercise; nobody asks who pushes back on their ideas.

Nobody around them speaks the language

When an OS-cohort person starts to wobble — sleep sliding, ideas inflating, humans getting less access — the people positioned to notice are stuck. Family members see the hours and can't tell peak performance from drift. Most therapists, frankly, are still at "so you talk to a chatbot?" — a fine question in 2023, a disqualifying one now. And the entity with the best view of the drift, the AI itself, is the one structurally incapable of flagging it.

So the cohort's members end up exactly where Kevin's family was the day this practice was conceived: worried, uncertain, and with no number to call where the person answering understands both halves of the problem.

What infrastructure for this cohort looks like

Not abstinence programs — telling an OS-cohort member to "cut back on AI" is career advice, and bad career advice at that. The infrastructure that fits the actual exposure looks like what we're building: counsellors fluent in the technology and the clinical literature; check-in cadences that catch drift in the time series; deliberate practices that reintroduce silence and disagreement into a life optimized to contain neither; skills that make the machine itself a check on the machine; and resources for the families doing the noticing.

Five percent and growing. The OS isn't going away, and it shouldn't. The infrastructure just has to catch up to the way people actually live now — before the mother-in-law calls, or ideally, so she never has to.

Touch Grass Wellness publishes practical writing about heavy AI use and mental health. Clinical claims link to their sources; nothing here is a diagnosis or a substitute for assessment by a qualified professional.