Fetching from the wire…
01
02
03
04
05
06
07
08
Top 5 · 2026-05-16 · source-backed
The HashiCorp founder posted something that's been rattling around my head all day. He believes there are "entire companies right now under heavy AI psychosis" where rational conversation about what AI can and can't do has become impossible. The post hit 1,812 points and 1,014 comments on Hacker News, making it the highest-engagement AI discussion this week.
What makes this land differently than the usual "AI is overhyped" take is Hashimoto's framework. He draws a direct parallel to the early cloud adoption era, specifically the MTBF vs MTTR reckoning. Back then, companies that couldn't have honest conversations about failure rates shipped brittle systems and paid for it later. Now he's seeing the same pattern: organizations embracing "it's fine to ship bugs because agents will fix them" without ever testing that assumption.
The timing is brutal. Same week, Amazon workers are openly admitting to "tokenmaxxing", making up tasks to hit mandatory 80%+ AI usage quotas. The internal tool MeshClaw is being used to automate unnecessary code deployments and emails, inflating usage scores with real production actions. This is Goodhart's Law playing out in real time at one of the world's largest engineering orgs.
Then Mistral's CEO tells French Parliament that his engineers "no longer write a single line of code." And Airbnb announces 60% of new code is AI-generated with "one engineer doing the work of 20."
Here's what I think is actually happening: there's a spectrum between "AI is useless" and "AI replaces all coding." The psychosis isn't believing AI is powerful. It's losing the ability to distinguish between a 71% productivity gain (Stanford's number, which I'll get to) and a 100% replacement. That gap is where all the bugs live.
What builders should do: Hashimoto's MTBF/MTTR framework is the diagnostic. Ask yourself: does your org track AI-generated bug rates separately from human-written code? Can you have a conversation about where agents fail without someone treating it as anti-AI? If no, you're in the psychosis zone. The fix isn't less AI. It's better measurement.
Each link below shares sources, entities, or timing with this story.
Amazon invested in OpenAI / Shared entity: Same / Shared topic / What happened next / Tension
Linked by a graph relationship (Amazon invested in OpenAI); both cover Same; overlapping topics (agent, between, code).
Amazon supports Claude Code / Shared entity: Hacker News / Same source domain / Shared topic / What happened next
Linked by a graph relationship (Amazon supports Claude Code); both cover Hacker News; reported by the same outlet (news.ycombinator.com).
Amazon invested in OpenAI / Shared entity: Hacker News / Same source domain / Shared topic / What happened next
Linked by a graph relationship (Amazon invested in OpenAI); both cover Hacker News; reported by the same outlet (news.ycombinator.com).
Amazon supports Claude Code / Shared entity: Same / Shared topic / Earlier coverage
Linked by a graph relationship (Amazon supports Claude Code); both cover Same; overlapping topics (agent, company, engineer, have).
Amazon supports Claude Code / Shared entity: Hacker News / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (Amazon supports Claude Code); both cover Hacker News; reported by the same outlet (news.ycombinator.com).
Amazon invested in OpenAI / Shared entity: Same / Shared topic / What happened next
Linked by a graph relationship (Amazon invested in OpenAI); both cover Same; overlapping topics (agent, between, code).
Linked by a graph relationship (Amazon invested in OpenAI); both cover Same; overlapping topics (agent, company, engineer).
Amazon supports Claude Code / Shared entity: Hacker News / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Amazon supports Claude Code); both cover Hacker News; overlapping topics (agent, code).