Hashimoto, who built HashiCorp into a $5B+ infrastructure company and created Terraform, Vagrant, Vault, and Consul, says he 'strongly believes' entire companies are now operating under 'AI psychosis.' The deliberately clinical term implies organizations have lost contact with reality, restructuring hiring plans, product roadmaps, and strategies around AI capabilities that remain unreliable and poorly understood by decision-makers.
The editorial argues that Hashimoto's framing captures something prior AI-hype complaints missed: naming it as a systemic organizational pathology rather than an individual opinion. Companies are laying off experienced engineers claiming AI will fill the gap, then quietly rehiring months later when AI-augmented skeleton crews can't ship — the economics of replacing senior engineering judgment with API calls remain undefined because that judgment can't actually be replaced yet.
The submission scored 1292 points with 637 comments on Hacker News, placing it in the top fraction of a percent of all submissions. This extraordinary engagement signals that thousands of practicing engineers saw their own workplace dynamics reflected in Hashimoto's diagnosis, lending grassroots validation to the 'AI psychosis' framing.
Mitchell Hashimoto — the engineer who built HashiCorp into a $5B+ infrastructure company, created Terraform, Vagrant, Vault, and Consul, and is now building the Ghostty terminal emulator — posted a blunt assessment on X: he "strongly believes" there are entire companies now operating under what he calls "AI psychosis."
The term is deliberately clinical. Psychosis implies a break from reality — and that's precisely what Hashimoto is diagnosing: organizations making strategic decisions based on what AI *might* do rather than what it demonstrably *does*. The post scored 1292 on Hacker News, placing it in the top fraction of a percent of all submissions — a signal that thousands of working engineers recognized their own workplace in the description.
Hashimoto isn't an AI skeptic or a Luddite. He's an infrastructure engineer who thinks in systems. When someone with his track record — shipping tools used by millions of developers daily — says the industry has lost contact with reality, practitioners listen.
The "AI psychosis" framing captures something that individual complaints about AI hype haven't: it names the phenomenon as an organizational pathology, not an individual opinion. This isn't one PM adding a chatbot nobody asked for. It's entire company strategies — hiring plans, product roadmaps, M&A decisions, layoff rationales — being restructured around capabilities that remain unreliable, expensive, and poorly understood by the people making those decisions.
Consider the pattern visible across the industry in 2025-2026:
Staffing psychosis. Companies laying off experienced engineers and claiming AI will fill the gap — then quietly rehiring six months later when the AI-augmented skeleton crew can't ship. The math rarely works: a senior engineer costs $300-400K fully loaded. The Claude/GPT API costs to replace their judgment on ambiguous problems are... undefined, because you can't actually replace that judgment yet.
Product psychosis. Features being shipped not because users requested them, but because "AI-powered" appears in the investor update. The result: products that were coherent become incoherent. Users who chose your tool for its specific, well-designed workflow now get a chatbot bolted onto the side that hallucinates their configuration files.
Architecture psychosis. Companies committing to massive GPU infrastructure spend — often 10-100x their previous compute budgets — based on projections that assume current AI capability curves continue indefinitely. When the curves flatten (as they historically do for every technology), the CapEx is already sunk.
The Hacker News discussion reflected practitioners at every level recognizing these patterns. Comments described being asked to "add AI" to systems where the failure mode of hallucination is catastrophic — medical records, financial calculations, infrastructure provisioning. The common thread: the people making the decisions don't write code, don't understand the failure modes, and interpret engineer pushback as resistance to change rather than risk assessment.
Not every aggressive AI bet is psychosis. Some companies *are* building genuine competitive advantages with AI — Cursor for code editing, Harvey for legal research, Midjourney for image generation. The difference is specificity: these companies built AI into a narrow, well-defined workflow where the failure modes are understood and the value proposition is concrete, not "we added AI to everything."
The counter-argument from the investment side: in platform shifts, being early and wrong on timing is often better than being late and right. Amazon lost money for a decade. Mobile-first companies looked insane in 2008. Maybe some of today's "psychotic" AI bets will look visionary in 2030.
But there's a crucial asymmetry. When Amazon bet on e-commerce, they were making a bet on *logistics* — a domain with measurable, improvable metrics. When a company guts its QA team because "AI will handle testing," they're making a bet on a technology whose failure mode is *silent wrongness*. The downside isn't "we spent too much on warehouses." It's "we shipped bugs we can't even detect."
If you're an IC or engineering leader inside one of these organizations, Hashimoto's framing gives you vocabulary for what you're experiencing. Here's how to use it:
Demand specificity. When leadership says "we're going AI-first," ask: "AI-first for which workflow, with which failure modes accepted, measured by which metric, with what fallback?" Psychosis thrives on vagueness. Specificity is the antipsychotic.
Document the reversibility. If your org is making AI-driven architectural decisions, ensure they're reversible. Don't delete the traditional code path. Don't lay off the person who understands the non-AI system. The cost of maintaining optionality is tiny compared to the cost of rebuilding institutional knowledge.
Calibrate your own career decisions. If your company is making decisions that seem disconnected from engineering reality, that's signal. Companies in psychotic episodes often have excellent stock prices (the market rewards narrative) but terrible engineering cultures (the people who see reality leave first). The ones who stay inherit the cleanup.
Hashimoto's post will age in one of two ways: either as a prescient diagnosis of a bubble-era pathology that self-corrected when the money got tight, or as the kind of conservative take that looks foolish after a technological step-function. History suggests it's the former — not because AI isn't transformative, but because organizational psychosis never produces good engineering, regardless of what technology triggered it. The companies that win with AI will be the ones that treated it as a tool to be evaluated, not a religion to be adopted. Those companies probably aren't tweeting about their AI strategy. They're too busy shipping.
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