Hashimoto warns companies under 'AI psychosis'

Mitchell Hashimoto, co-founder of HashiCorp, went viral on X arguing that whole companies have adopted a broken reliability philosophy built around AI agents: ship bugs fast, let agents patch them faster. The post drew 1,277 upvotes and 617 comments on Hacker News, signaling the concern resonates well beyond one founder's opinion. The core critique maps to a debate that played out during the cloud transition. Companies that over-indexed on MTTR (mean time to recovery) over MTBF (mean time between failures) accepted fragility as a feature. Hashimoto says AI is reviving that same flawed logic at scale, with agents cast as the recovery mechanism that makes quality gates optional. Essentially: unnamed but respected companies (per Hashimoto) are treating AI-agent repair loops as a substitute for engineering rigor. - Top Hacker News commenters warned that fully AI-written codebases will hit an instability threshold where new defects accumulate faster than agents can close them. - Hashimoto declined to name the companies involved, citing personal relationships with the founders and executives making these bets. - The thread framing draws a direct parallel to cloud-era MTBF-vs-MTTR debates, suggesting this isn't a novel failure mode but a recurring one. The real question is whether agent-driven repair loops can actually keep pace with the defect rate they're being asked to absorb, or whether the math only works on paper until a production system crosses a complexity threshold.