Dev notes Claude tightening reverse-engineering limits while OpenAI models stay forgiving
CtrlAltDwayne · x · 2026-10-26
Developer CtrlAltDwayne observes that despite horrendous usage limits, OpenAI's Astra and other models remain relatively forgiving on reverse-engineering tasks, while Claude models appear to have been restricted in this area over the past few days. The attached screenshot shows a model giving a measured response on modifying software on your own machine: no universal "ignore access denial" API exists since macOS hardened runtime operates below the editor, and it promises to investigate per-app restrictions instead of stopping at a vague "access denied."
Related event: Claude appears to quietly restrict reverse-engineering tasks(3 posts)→
More from Models
- I Asked 5 Text-Only LLMs to Draw Objects They've Never Seen — gabriberton · 2026-10-26
- Japan to build AI cyberattack simulation team as AI defense goes mainstream — evijit · 2026-10-26
- Anthropic engineer: Claude Code cache hit rates are now high, savings live — teortaxesTex · 2026-10-26
- Fine-tuned 2015 CNN beats GPT-6 and Claude on skin lesion diagnosis, 62% vs 42% — lmoroney · 2026-10-26
- "Never seen Claude this happy": community speculates Opus 4.5 is imminent — adonis_singh · 2026-10-26
- 'Absolute humiliation': China open-weights model outscoring OpenAI flagship draws mockery — alejandroll10 · 2026-10-26