Ethan Mollick on steering long-running agents: when to instruct, queue or fork
emollick · x · 2026-09-10
Ethan Mollick argues the critical factor in using agents in Codex and Code is deciding when to add instructions mid-task (risks anchoring) versus queueing or forking conversations (risks wasted or incorrect output). You can't be in the loop for complex long-running tasks, but you can oversee the loop — if you're not steering long-running agentic work, you're not managing agents enough or building interim reporting.
Related event: Ethan Mollick on 'driving' long-running agents like Codex(4 posts)→
More from coding & agent
- Amp adds context analysis to expose MCP tool and AGENTS.md bloat in threads — HankYeomans · 2026-09-10
- Hermes Agent adds live subagent monitoring with manual steer and stop controls — NousResearch · 2026-09-10
- Talkthrough MCP lets Claude Code search errors and quotes in videos, open source under MIT — phdptsd · 2026-09-10
- Stageflow: an OSS multi-stage agent workflow runtime seeks feedback on viability — tejasghutukade · 2026-09-10
- STM32 Motor Board Running in Under an Hour with Copperkit — debreuil · 2026-09-10
- Cognition launches SWE-2: frontier-level coding performance at up to 70% lower cost — silasalberti · 2026-09-10