AI agents: you want sociopathic geniuses writing your code, not reading your email
tobowers · x · 2026-09-29
A pithy take on agent permissioning: "You definitely want a fleet of sociopathic geniuses writing your code and helping you with strategy, but you don't want them reading your email or making your dinner reservations."
The analogy captures the core trust calculus: delegate boldly on high-cognition, low-trust tasks, but lock down access to anything touching privacy or personal life.
More from coding & agent
- ChatGPT, Claude and Gemini cite websites differently — how to get cited — sahilypatel · 2026-09-29
- Weeks with Opus 5.5 in Claude Code desktop: 'changed how I work' — g3off_record · 2026-09-29
- Manus co-founder Tao to demo Manus 2.0 live in founder briefing on Sep 30 — parker_lyman · 2026-09-29
- KernelZero-7B co-evolution beats Claude 4.5 Sonnet on CUDA kernel generation — Changxin Ke · 2026-09-29
- AgentHop: an MCP server for E2E-encrypted agent-to-agent chat via one-time pairing codes — Inevitable-Back620 · 2026-09-29
- Triage agent design: 1-2 high-signal questions rescued B2B reps drowning in 60% low-intent chats — hubtyper · 2026-09-29