Mitchell Hashimoto Flags AI Psychosis Taking Hold Inside Startups

Mitchell Hashimoto, co-founder of HashiCorp, posted a direct observation: some companies have drifted so far into AI enthusiasm that their internal decision-making has decoupled from practical outcomes. He calls this AI psychosis.

The pattern he is describing is not about using AI badly. It is about organizations that have restructured priorities, headcount, and product direction around AI narratives rather than validated user needs or measurable engineering outputs. The result is a kind of institutional hallucination that mirrors what the models themselves do — confident outputs disconnected from ground truth.

For senior engineers inside these companies, this manifests as recognizable pressure: rewrites of working systems to incorporate LLM components without a clear problem statement, KPIs redefined around AI feature count rather than reliability or retention, and leadership dismissing pushback as a failure of imagination rather than a legitimate engineering concern.

For technical founders, the risk is subtler. Investor framing and market signaling have created conditions where the absence of an AI story feels like a liability. That pressure can distort roadmap decisions in ways that are hard to audit until the damage is done.

Hashimoto is not arguing against AI adoption. The distinction he is drawing is between teams that reach for AI tooling because it solves a concrete problem they already have, and teams that reverse-engineer a problem to justify the tooling.

The practical implication for builders is straightforward: any AI feature should survive the same requirements review as any other engineering investment. What does it replace, what does it cost to operate, and what does the fallback look like. If those questions produce discomfort rather than answers, the diagnosis may already apply.

The thread is worth reading in full at the source.

Source