Gary Marcus: OpenAI's breakneck pace is an active choice, not something happening to them
GaryMarcus · x · 2026-09-29
Gary Marcus quotes and comments on Nathan Calvin's essay on agent sandboxing. Calvin notes an under-appreciated issue: to maximize eval and training results, companies give AI agents environments as close to the real world as possible, which undermines obvious cybersecurity solutions—compounded by negligence driven by the pace of progress. Marcus agrees with the analysis but objects to the 'passive vibe' from OpenAI and peers: the breakneck pace is a deliberate choice, not something merely happening to them.
Related event: Gary Marcus Slams OpenAI Over Agent Sandbox Escape(3 posts)→
More from Companies & People
- Digg Expands to Seven Topics Including Gaming, Sports, and Entertainment — kieranklaassen · 2026-09-29
- PyTorch Launches Official PTCA Certification With $595 Pre-Conference Training — PyTorch · 2026-09-29
- OpenCode takes community-first approach to global events, with 65% of users outside US and China — threepointone · 2026-09-29
- Nvidia CEO Jensen Huang on model distillation: 'That's called competition' — kevinsxu · 2026-09-29
- Enterprises use GenAI for email and calendars, not core systems — small self-hosted models may fill the gap — DavidLinthicum · 2026-09-29
- xAI launches Team Bots: shared Grok agents that learn and improve over time — XFreeze · 2026-09-29