A software engineer with a long public track record posted something blunt to Hacker News this week: he believes entire companies are currently operating under what he called "AI psychosis" — organizations where AI-generated output has become so embedded in decision-making that leadership has lost the ability to evaluate whether any of it is grounded. The thread lit up. The replies were not mostly dismissive.

That reaction is a data point worth sitting with.

What "AI psychosis" actually looks like from the outside

The phrase is deliberately provocative, but the underlying pattern is recognizable. A company deploys AI tools across forecasting, hiring, communications, and customer service. Each layer of output trains the next layer of decisions. Managers stop asking whether the model's recommendations make sense because the recommendations arrive confidently formatted and no one has time to audit them. The organization starts optimizing for what the AI can measure, not for what's actually true.

This isn't a future scenario. It's a failure mode with a historical analog: the spreadsheet-driven management culture of the 1980s and 1990s, where the precision of the number became confused with the accuracy of the number. AI makes that confusion faster, more fluent, and harder to spot from inside.

The difference now is scale and speed. A spreadsheet error took a quarter to damage a supply chain. An AI feedback loop baked into procurement, staffing, and logistics decisions can compound across weeks.

Why a household should care

Most families don't think of themselves as downstream of any particular company's internal software culture. But if your employer uses AI-generated performance metrics to determine layoffs, you are. If your health insurer uses AI claim review to approve or deny treatment, you are. If the company managing your 401(k) custodian uses AI-generated risk models to rebalance during a volatile week, you are.

Recent BLS data on layoff patterns shows that white-collar job displacement is no longer concentrated in any single sector. Software, finance, legal services, media, and healthcare administration have all seen significant reductions in the past 18 months. Many of those reductions were announced with language about "efficiency" and "AI-augmented workflows."

The household-level risk isn't that AI is malevolent. It's that organizations optimizing poorly with AI can make large, fast, irreversible decisions — and the people affected get a form letter.

What we'd actually do

Build a three-month expense buffer before you need it. Not a preparedness lecture — a math problem. If your household could absorb 90 days without primary income, most sudden layoffs become a disruption rather than a crisis. That number is different for every family; the point is to know yours and work toward it. Even a 60-day buffer changes the emotional math of a job search.

Know your industry's AI exposure, honestly. Not every role is equally at risk. Administrative work with clear, auditable outputs is more exposed than roles requiring physical presence, long client relationships, or genuine judgment under ambiguity. Understanding where your work sits on that spectrum is not about panic — it's about career planning with accurate inputs. The O*NET occupational database is free, regularly updated, and shows task-level breakdowns for most jobs.

Diversify income before a crisis, not during one. One consulting client, one freelance relationship, one skill rented to a neighbor — any income thread outside your primary employer reduces your household's single-point-of-failure risk. This doesn't require a side hustle empire. It requires starting before you need it.

Get a paper copy of your benefits documentation. If your employer's HR runs on AI-managed portals and something goes wrong administratively — a systems error, a contract dispute, an unexplained status change — knowing your exact benefits, vesting schedule, and coverage terms from a document you control matters. Download and print these annually.

Pay attention to organizational signals. Mass internal AI tool rollouts, sudden changes in how your manager talks about "efficiency," new performance tracking systems you don't recognize — these are not definitive signs of anything, but they're worth noticing. Workers who see layoffs coming six weeks out make better decisions than those who are blindsided.

The bigger picture

The problem named in that Hacker News thread isn't AI itself. It's organizations that move faster than their judgment can follow. That has happened with every powerful new tool in the past century, and the households that came through those transitions best were the ones with financial margin, portable skills, and more than one thread holding them up.

Durability isn't about predicting which company goes sideways. It's about being able to absorb surprise, whatever its source.