Entire Companies Are Under AI Psychosis — And That's a Problem for No-Code

Mitchell Hashimoto knows a thing or two about building things people actually use. He co-founded HashiCorp. He built Vagrant, Terraform, Vault. He's currently building Ghostty, the terminal emulator that developers have been obsessing over. He is not, by any stretch, a hype merchant.

So when he tweeted "I strongly believe there are entire companies right now under heavy AI psychosis and its impossible to have rational conversations about it with them," people paid attention. 2,069 points on Hacker News. 1,225 comments. It's the kind of number you only get when you name something a lot of people have been thinking but couldn't quite articulate.

"AI psychosis" is a pretty good name for it. You've seen it. The Slack message from leadership announcing a new AI initiative with no stated purpose. The all-hands where someone asks how many employees are using Copilot this week, as if it's a pulse oximeter reading. The vendor pitches that now have an AI slide wedged in whether or not the product actually does anything with a language model. Everyone performing belief in a thing nobody can define.

The evidence that this is real, and not just vibes, is piling up fast.

Amazon is the most damning case study. The company has been tracking something called "AI token consumption" and pressuring teams to hit usage quotas. The response from workers was immediate and predictable: they started making things up. Amazon employees built AI agents whose sole function was to burn tokens and look busy on the dashboard. They used the company's internal "MeshClaw" tool to automate nonsense tasks, not because the tasks needed doing, but because the metric demanded it. People on forums have started calling it "tokenmaxxing."

Let that sink in. One of the most sophisticated technology companies on earth is running an incentive programme that causes its own employees to generate fake work so middle management can present a green chart at the next quarterly review. This is theatre. It's not productivity. It's compliance dressed up as innovation, and every middle manager in the chain knows it.

The Fortune survey from April makes the problem structural, not anecdotal. Eighty per cent of white-collar workers are refusing AI adoption mandates. Not secretly sabotaging them. Just quietly doing the work manually, the way they always have, while leadership stares at dashboards showing adoption climbing. Fortune called it FOBO: Fear of Becoming Obsolete. Employees aren't resisting AI because they're Luddites. They're resisting because nobody has explained what the AI is supposed to replace, and the unspoken answer, they suspect, is them.

So you get a two-tier reality. Upstairs: "our AI transformation is accelerating." Downstairs: people doing spreadsheets by hand and inventing fake prompts so the graph goes up.

This is the corporate AI backlash. And it's happening at the exact moment the most consequential transformation in how software gets built is unfolding on the ground, inside the same companies, largely ignored by the people writing the mandates.

If you work in or around no-code, you should be paying close attention to this moment. Because it's a window.

Why the Backlash Creates an Opening

The AI psychosis Hashimoto is describing has a specific shape. It's top-down. It's metric-driven. It starts with "we need to use more AI" rather than "here's a problem we need to solve." The metric becomes the goal and the goal displaces the work.

No-code has always worked the other way. It starts with someone who needs to build something, and it gives them the tools to build it faster. Nobody adopts Bubble or Glide or Bolt.new because there's a quota. They adopt them because they need a working app and they need it this week. The productivity gain is self-evident. You don't need a dashboard to prove it. You just look at what shipped.

This is the distinction that matters. There's a wide gulf between "AI adoption" as a corporate KPI and "using AI-powered tools to actually get things done." The first is measurable in tokens. The second is measurable in working software. The companies trapped in psychosis are measuring the wrong thing, and their employees know it. The companies that get this right are measuring output, not input, and they're the ones quietly shipping.

I think the no-code world has been strangely quiet about this dynamic. For years we've had to defend ourselves against the accusation that visual development tools aren't "real" engineering. Now the biggest companies on earth are burning millions on AI licences so their employees can generate fake work to satisfy a metric nobody understands. It's hard to look at that and not feel like the no-code approach — pick the right tool for the job, measure by what ships — has been right all along.

This isn't an argument for smugness, though. There's a danger here too.

The Indiscriminate Backlash

When the AI psychosis bubble pops, as it will, the reaction won't be careful and surgical. It'll be a machete. CFOs will look at AI line items and cut them across the board, without distinguishing between "we paid for ChatGPT Enterprise licences nobody uses" and "we used Cursor to ship three internal tools this quarter." The people who actually integrated AI sensibly into their workflows will get caught in the same blast radius as the people who bought AI because their board asked about it.

We saw this pattern before, in the no-code 1.0 era. Companies went all-in on "citizen developers," bought expensive enterprise licences for platforms nobody was trained to use, then declared no-code a failure when the dashboards didn't turn green. The tools weren't the problem. The rollout model was. And when the backlash came, it took down good projects alongside the bad ones.

The "ban AI" reaction that's brewing in response to the psychosis is just as misguided as the original mandate. It replaces one cargo cult with another. The question isn't "should we use AI" or "should we ban AI." The question is: does this tool help someone ship something real, faster? If yes, use it. If no, don't. That's it. That's the entire decision framework.

The Takeaway

For no-code builders, the opportunity here is straightforward. Be the person in your company who can point to a thing that exists because you used the right tools. The dashboard that went live in three days. The internal app that replaced a spreadsheet. The customer portal that would have taken six months of traditional development. When the AI psychosis hangover arrives, the people who can say "I shipped this" are going to look very different from the people who can only say "we hit our token quota."

Hashimoto's tweet resonated because it named something true. Companies are acting irrationally about AI, and the rational conversations are getting harder to have. But the solution isn't to opt out of the conversation. It's to change it. Talk about what got built, not what got prompted. Talk about time saved, not tokens consumed.

That's a conversation no-code builders know how to have. We've been having it for years. The AI psychosis era just makes it more urgent.