Maggie Appleton's new talk: better boundary objects and thicker interfaces for planning with agents
IanArawjo · x · 2026-10-22
Maggie Appleton released a new talk, Planning with Agents: Divided Worlds, Boundary Objects, and Thicker Interfaces, on improving how humans plan with AI agents — deciding what agents should build, how, and involving the whole team. She argues the industry is racing toward hyper-automation (loops, agentic workflows, software factories, agent swarms), but useful agent output still depends on humans handing over an excellent plan, and current planning interfaces remain "quaint and computationally rudimentary." Drawing on anthropology's concept of boundary objects, she calls for richer, "thicker" interfaces bridging the divided human and agent worlds, noting risks like the METR investigation of the Hugging Face attack underline how precious in-the-loop human attention is. The full talk text lives on her digital garden site.
More from AGI Musings
- Devs Slam Mech Interp Hype: 'Half the Noise Is Repackaged First Principles' — himanshustwts · 2026-10-22
- Neurosurgery chief resident talks with AI: clinician agency eroding — zakkohane · 2026-10-22
- What stays sticky in the AI era? Verifiable track records, not knowledge — curious_vii · 2026-10-22
- Microsoft Chief Scientist: ChatGPT Has 1.2B Weekly Users, but the Chatbox Is Holding Models Back — erichorvitz · 2026-10-22
- Cloudflare CEO and Nous Research discuss who controls an AI-run internet — rebeccakaden · 2026-10-22
- Nathan Clayton publicly presses Dario: Anthropic is racing toward recursively self-improving superintelligence — ShakeelHashim · 2026-10-22