Hashimoto uses the clinical term 'psychosis' — implying detachment from reality — to describe companies making strategic decisions based on unfalsifiable beliefs about AI rather than product needs or user demands. His credibility stems from building infrastructure tooling used by millions and taking HashiCorp from founding through IPO.
Posted Hashimoto's tweet to Hacker News where it received 1,921 points and 1,118 comments, indicating massive community resonance with the 'AI psychosis' framing. The extraordinary engagement suggests this articulated something many developers had felt but couldn't name.
The editorial argues this differs from previous hype cycles (blockchain, microservices, serverless) because AI narrative has captured executive decision-making at unprecedented speed while the technology's actual capabilities remain poorly understood by those making the decisions. The gap between narrative and technical reality is wider than in past cycles.
Identifies a specific anti-pattern: engineering teams spend months integrating LLM calls into workflows where deterministic logic would be more reliable, faster, and cheaper. The resulting AI features ship with quality guardrails that effectively neuter the output, delivering a slightly worse product with a chatbot bolted on — but the stock price goes up on announcement, perpetuating the cycle.
Mitchell Hashimoto — co-founder of HashiCorp, creator of Terraform, Vagrant, Packer, and currently building the Ghostty terminal emulator — posted a blunt assessment on Twitter: he "strongly believes" there are entire companies now operating under what he calls "AI psychosis." The tweet hit Hacker News and racked up over 1,900 points, making it one of the highest-scored posts of the week.
Hashimoto isn't a random commentator. He built infrastructure tooling used by millions of developers, ran a company from founding through IPO, and is now deep in systems-level programming on Ghostty. When someone with that track record uses the word "psychosis" — a clinical term implying detachment from reality — the developer community pays attention.
The term resonated because it names something many practitioners have felt but struggled to articulate: organizations making technical and strategic decisions not based on what their product needs or what their users want, but based on an unfalsifiable belief that AI transforms everything it touches.
The concept of "AI psychosis" is distinct from normal hype cycles. Every technology wave produces irrational exuberance — we saw it with blockchain, with microservices, with serverless. What makes this different, according to the emerging practitioner consensus, is the speed at which AI narrative has captured decision-making at the C-suite level while the technology's actual capabilities remain poorly understood by those making the decisions.
Consider the pattern: a company announces an "AI-powered" version of their product. The engineering team spends months integrating LLM calls into workflows where deterministic logic would be more reliable, faster, and cheaper. The AI features ship with quality guardrails that effectively neuter the AI's output. Users get a slightly worse product with a chatbot bolted on. But the stock price went up on the announcement, so the cycle continues.
This isn't hypothetical. We've seen search engines add AI summaries that hallucinate answers to medical questions. We've seen code editors where the AI suggestions introduce security vulnerabilities faster than developers can review them. We've seen companies lay off QA teams because "AI will handle testing" — then ship more bugs than ever. The psychosis framing suggests these aren't rational cost-benefit miscalculations; they're symptoms of an organization that has lost the ability to evaluate evidence.
The Hacker News discussion — massive by any standard at nearly 2,000 points — revealed several distinct camps. One group sees the problem as primarily a management failure: executives who don't understand the technology making bets based on competitor pressure and investor expectations. Another camp argues it's more structural: once an organization publicly commits to an AI strategy, admitting it doesn't work becomes career-ending for everyone involved. A third perspective suggests the psychosis is rational at the individual level — if the market rewards AI announcements regardless of product quality, the incentive structure makes delusion profitable.
For senior engineers, the "AI psychosis" framing clarifies a common workplace frustration. You're in a planning meeting. Someone proposes using an LLM for a task that's essentially string parsing with known formats. You suggest regex or a simple state machine — it'll be faster, deterministic, testable, and cost nothing to run. The response: "But we need to be an AI-first company."
That's the psychosis in action. The technology choice isn't being made on engineering merit. It's being made on narrative merit. And the engineer who pushes back risks being labeled "not aligned with the company's AI strategy" — which in 2026 is career poison at many organizations.
The practical advice emerging from the community: document everything. When you're overruled on a technical decision for narrative reasons, write it down. Note the alternative you proposed, the expected failure modes of the AI approach, and the metrics you'd use to evaluate both. When the AI approach inevitably hits the problems you predicted, you'll have the receipts — and more importantly, you'll have a concrete proposal ready to ship.
Hashimoto's observation has immediate implications for how you evaluate tools and vendors. If a product's primary marketing message is "AI-powered" rather than describing what problem it solves and how well it solves it, treat that as a yellow flag. The best AI-enhanced tools — Copilot at its best, Cursor when it works, specialized models for specific domains — lead with the outcome, not the mechanism.
For engineering leaders: the antidote to AI psychosis is empirical discipline. Before adopting any AI feature or tool, define the success metric, the baseline without AI, and the minimum improvement threshold that justifies the added complexity and cost. If you can't measure whether it's working, you can't distinguish strategy from psychosis.
For individual contributors: learn to distinguish between "this AI tool genuinely makes me faster" and "I'm using this AI tool because my company expects me to." The first is engineering. The second is compliance theater. Both exist simultaneously in most organizations right now, and being honest about which is which will serve you better than pretending everything AI touches turns to gold.
The psychosis framing suggests this is a temporary condition — psychosis, by definition, can be treated. The cure is usually contact with reality, which in software means shipping products and measuring outcomes. Companies running AI features will eventually face the spreadsheet: does this feature retain users, reduce costs, or unlock new revenue? Those that pass survive. Those that don't will quietly deprecate their AI features while the press release archive gathers dust. The question isn't whether the correction comes, but how much engineering talent and product quality gets burned before it arrives.
I feel in a really weird position where I both really dislike what AI is doing to the experience and practice of writing code, to the point where I want a job doing literally anything else besides using the computer, but also think that these tools are extremely powerful and only getting better.I th
Maybe this is what will turn software engineering into an Engineering field.Right know, prompters are setting up whole company infrastructure. I personally know one. He migrated the companies database to a newer Postgres version. He was successful in the end, but I was gnawing my teeth when he descr
Recently I had a request come through to allow finance analysts to vibe code their apps. During a discussion one of the finance managers let the cat out of the bag. Turns out our CFO had met fellow CFOs at a get together. They talked about how each of them were using AI. Our CFO was lagging behind a
I think AI rescue consulting is going to be come a significant mode of high value consulting, similar to specialists who come in to try and deal with a security breach or do data recovery.Purely AI written systems will scale to a point of complexity that no human can ever understand and the defect c
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
I’m at a FAANG and we have $300/day token quota. Personally I don’t use that much of it but management is pushing really hard for it. “the quota has been raised for a reason, use it”. Any task: “have you tried working on it with Claude?”. Every meeting “now engineer x and y will show you what h