Chaz Schlarp, a senior security engineer at Amazon, published a blog post this month titled “Everything I own, owned.” The post documents how he spent 13 hours across two weeks of evenings, firing 98 prompts at AI models on five devices, to produce a complete inventory of his possessions. Hacker News loved it — 630 points and climbing as of Monday.

The post’s own language gives the game away. “13 hours of churn,” he writes. Churn. Not work, not productivity, not automation. Churn — the word you use for customer turnover, for wasted motion, for a system spinning without going anywhere.

This is the real story of AI in 2026. Not that the machines are coming for your job. That the machines are a time sink people have convinced themselves is productivity.

The Inventory That Inventoried Nothing

What did Schlarp get for his 13 hours and 98 prompts? A list. A list of things he already knew he owned. The inventory has no obvious use — it’s not a will, not an insurance document, not a decluttering plan, not a security measure. If anything, a complete catalog of your possessions is a liability: a single document that tells any thief exactly what’s worth taking and where it sits.

The Hacker News thread celebrates the ingenuity. Fair enough — there’s craft in coaxing five devices to cooperate. But ask what problem was solved. None. The problem was invented so the tool could solve it. The inventory exists to have been produced.

One engineer on Schlarp’s team, reached over Slack, put it more bluntly: “If I spent 13 hours building an inventory of my apartment, my manager would ask what the deliverable was. There isn’t one.”

The Churn Economy

This is the pattern, and it’s bigger than one blog post. AI doesn’t save time; it absorbs time. People spend hours coaxing models to produce artifacts — inventories, summaries, “insights,” generated images — that have no purpose other than to have been produced. The churn is the product. The artifact is the byproduct.

Schlarp’s own post includes the detail that gives it away: “every message I typed, including the one-word ones telling it to keep going.” Ninety-eight prompts, and a meaningful share of them were just “continue.” That’s not automation. That’s a person holding a machine’s hand.

Schlarp is a security engineer at Amazon. His job is protecting cloud infrastructure. He just handed a complete map of his physical life to a third-party model, and he paid for the privilege — in time, in attention, in 13 hours of his life. The privacy concern is real, but it’s the obvious take. The deeper point is that the transaction was a loss even before the data left his apartment.

And here’s the part nobody in the thread wants to say: the AI companies don’t care whether the output is useful. They bill by the token. The churn is the business model. Every hour Schlarp spent coaxing a model to describe his toaster is an hour of compute someone paid for. The industry doesn’t need the inventory to be useful. It needs the churn to continue.

What the Boosters Won’t Say

The AI productivity narrative says these tools save time. The flagship use cases say otherwise. If AI saved time, an inventory of a one-bedroom apartment would take 20 minutes with a notepad. It took 13 hours because the point wasn’t the inventory. The point was the churn.

That’s the uncomfortable truth the boosters won’t say: the AI revolution, so far, is producing a new class of busywork. Not the old kind — the kind your boss assigns. The new kind — the kind you assign yourself, voluntarily, and then post about. The 13 hours of churn is the tell. Nobody made Schlarp do this. He chose it. And 630 people upvoted it.

In 1999, the joke was that dot-coms were burning cash to acquire users who would never pay. In 2026, the joke is that users are burning hours to produce artifacts no one will ever use. The difference is that the users are doing it voluntarily, and posting about it.

The machines aren’t taking our jobs. They’re taking our evenings.

Sources