AI models still can't predict shell and OS nuances, dev shows with pkill example
blixt · x · 2026-09-27
Developer blixt notes that after all this time, AI models still fail to predict the nuances of shells and operating systems. His example: what does pkill -f "codex app-server" -P 1 actually do on macOS? The gap between model prediction and real system behavior highlights how weak models remain on process matching and OS-level details.
Related event: Developer says AI models still can't predict macOS pkill behavior(2 posts)→
More from Models
- Claude's Snarky Tone and Mysterious "For Anyone Reading Along" Baffle Users — MrsChatGPT4o · 2026-09-28
- Matt Shumer: Opus 5.5 Added Controller Support to His Product — mattshumer_ · 2026-09-28
- Grok Bot already works in Teslas via the built-in Grok app, full release due this summer — yunta_tsai · 2026-09-28
- Claude Opus 5.5 reportedly balks at a population ethics problem, researcher suspects 'repugnant conclusion' aversion — anderssandberg · 2026-09-28
- Xiaomi releases MiMo-V2.6-Flash-MOPD on Hugging Face — Automatic-Arm8153 · 2026-09-28
- LLM confidently said Paris is Germany's capital—hallucination remains unsolved — ingliguori · 2026-09-28