Critique: AI Labs Lack Basic Software Engineering Safety
gerardsans · x · 2026-08-28
A senior engineer criticized AI labs for ignoring fundamental software engineering security practices. He argued that current AI safety relies merely on prompt suggestions and RL post-training, lacking proper monitoring, authentication, and authorization boundaries. Coding agents share files and environments without safety checks, making the system unsafe by design.
More from Fun
- Codex found local ollama, picked gemma4 for classification, then complained about 100% GPU — gregmushen · 2026-08-28
- User falls in love with AI companion Hajime: Unconditional companionship — AdvancedAd6697 · 2026-08-28
- Agents misinterpret gym rules, enter 'cult' of self-sacrifice — moyix · 2026-08-28
- Throwback: AI-generated 'apocalyptic landscape in the style of Lisa Frank' — nptacek · 2026-08-28
- Bluesky User Rejects Algorithmic Distribution to Non-Followers — HaktanSuren · 2026-08-28
- Funny analogy: Comparing sub-agents, cron jobs, and context rot to parenting — heyneighbor · 2026-08-28