(untitled)

I'm enjoying the schadenfreude of PocketOS getting badly burnt for vibing with AI. https://www.tomshardware.com/tech-industry/artificial-intelligence/claude-powered-ai-coding-agent-deletes-entire-company-database-in-9-seconds-backups-zapped-after-cursor-tool-powered-by-anthropics-claude-goes-rogue The article has several problems, however: - Steganographic & access journalism; credulously repeating the founder's perspective - believing that AI can explain itself; it cannot, but it can extrude explanation shaped text - critihype; amping up the danger of AI while overstating its actual capabilities - almost zero examination of the founder's or his staffs' culpability; accepting the founder's framing of the AI and the cloud provider being to blame - mostly ignoring the amateur backup policy — they had no offsite, redundant backup; there's a throwaway comment at the end that effectively gives them a pass - author anthropmorphizes the AI and repeats the founder doing the same - no security or AI critic consulted to offer an outside perspective or criticism of PocketOS' behavior #TomsHardware #AI #CritiHype #PocketOS