The phrase "AI psychosis" is being thrown around to describe executives who fall in love with AI demos they never actually touch. It came up on TechCrunch's Equity podcast, "Making sense of the debate over AI psychosis", where the hosts picked apart a claim from Box founder Aaron Levie that tech CEOs are "uniquely prone" to it.
I want to push past the clickbait framing. The interesting part isn't a psychological diagnosis of billionaires. It's a working rule for the rest of us: judge AI by what it does in your own hands, not by what a slide promises.
🔍 What "AI psychosis" actually means here
Strip away the drama and Levie's point is mundane and correct. A CEO who hasn't written code in years, or never did, looks at a productivity chart and concludes the machine has solved the problem. The person three layers down, who has to ship the thing, sees where the output breaks.
Levie isn't anti-AI. His complaint is about distance from the work. As he put it on the show:
"You can't just look at a slide and be like, 'Yes, incredible efficiency, let's go.'"
That gap between the slide and the shipped product is the whole story. It's not unique to CEOs either. Any of us can catch it.
Key takeaway: "AI psychosis" is just decision-making without hands-on evidence. The cure is cheap: use the tool yourself before you bet on it.
📊 The hype gap, in one table
The Equity discussion sat inside a wider backlash against AI being stuffed into products whether users asked for it or not. Two concrete signals came up:
| Signal | What happened | What it tells you |
|---|---|---|
| DuckDuckGo installs | Up roughly 30% as people reject AI-heavy search | Users will leave when AI degrades a product they trust |
| Google accuracy | Reported fumbling basics, including misspelling its own name | Shipping AI fast doesn't mean shipping it right |
One host summed up Google's bind:
"Google is chasing that thing it feels like it has to do to keep up, but it's messing with the thing that people attach to the brand the most."
Those are the only two data points I'll stand behind, because they're the only ones in the source. I'm not going to invent a benchmark to make the post look thorough. The pattern is enough: pressure to look AI-first is pulling companies away from what their users actually valued.
🛠️ Why Sri Lankan builders are on the right side of this
Here's the part that flips the story. The whole risk Levie describes comes from being far from the work. Most of us reading this are the opposite. If you're a student, a freelancer, or a three-person team in Colombo, you don't have a deck-reading layer between you and the keyboard. You are the implementation layer.
That's an advantage. You can ground-truth any AI claim the same afternoon you hear it:
- Pick the boring task you actually do — drafting a cover letter, cleaning a CSV, removing a background from a product photo.
- Run the AI tool on a real input, not a curated demo input.
- Check the output line by line. Where does it hallucinate? Where does it save you 20 minutes?
- Keep it only if step 3 was positive. Otherwise drop it, no sunk-cost guilt.
You don't need a budget to do this. Browser-based tools cost nothing and leak nothing. Our own free tools run AI client-side, so you can try something like the background remover on your own files and watch exactly where it succeeds and fails. That five-minute test teaches you more than any vendor slide.
Bottom line: the closer you are to the actual work, the harder you are to fool. Stay close to it on purpose.
💡 A small-team rule for AI decisions
You don't need a committee or a "transformation strategy." You need a habit. Before you adopt any AI tool or accept any AI productivity claim, ask three things:
- Did I run it on my own data? Not the demo. My messy, real input.
- Can I see where it fails? A tool you can't break is a tool you don't understand yet.
- Would I still use it if it weren't called AI? Strip the label. Is the result good?
If a tool passes those, it earns a place in your workflow. If it only passes the slide, it doesn't. This is the same discipline whether you're a solo freelancer or running Box. The difference is that you can apply it today, for free, without anyone's approval.
There's also a quieter lesson about layoffs. The Equity hosts noted AI adoption is real and already costing jobs, and that executive hype often runs ahead of ground truth. Both are true at once. The defence against being on the wrong end of that is to be the person who knows exactly what these tools can and can't do, because you tested them, not the person repeating a vendor's number.
🌐 What this means for you
The "AI psychosis" debate is a fancy label for an old failure: believing a claim you never checked. You sidestep it by staying close to the work and testing on real inputs.
- Don't argue about AI in the abstract. Open the tool, feed it a real task, judge the output.
- Treat free tiers as your test lab. Browser and free tools let you evaluate at zero cost and zero risk.
- Trust evidence over enthusiasm — yours and everyone else's, including mine.
The people most likely to fall for hype are the ones furthest from the keyboard. If you're reading this, you're probably not one of them. Stay that way.
Top comments (0)