If Your Agent Forgets Everything When the Task Ends, It's a Script — Give It a Persistent Machine
Suresh935 · reddit · 2026-11-17
The author argues the spin-up-clone-run-destroy sandbox model is wrong for agents working on a project for days: quality drops before the context window fills, and compaction loses what mattered. The converging fix is treating the filesystem as memory — progress files, codebase notes, a growing AGENTS.md, git history — and giving each agent its own long-lived machine, since dependencies, warm caches, code indexes and logged-in sessions can't be rebuilt cheaply (2FA logins can't be rebuilt at all). Practices: stop idle machines instead of deleting, snapshot before risky steps, fork for parallel approaches, scope credentials per agent. Downsides — drift, stale notes, growing attack surface — call for scheduled clean-baseline rebuilds, hard idle stops, and reviewing memory files like code. Tooling notes on Fly Sprites, exe.dev, Coder and boat.dev included.
More from coding & agent
- Zoho Desk launches Blended Agents: free to build, up to 30M free tokens monthly — shashib · 2026-11-17
- A prompt trick for delightful empty states: ask AI for SVG- animated easter eggs — vista8 · 2026-11-17
- Point an Agent at Agent Foundry's artifacts folder to debug time and cost — colinmcnamara · 2026-11-17
- Chollet: Advanced agentic coding is essentially machine learning, with ML's classic failure modes — fchollet · 2026-11-17
- United Bans Unauthorized AI Agents From Booking Flights, Southwest Embraces ChatGPT — Future_AGI · 2026-11-17
- DavidKPiano: no AI code review tool reviews the intent behind code — DavidKPiano · 2026-11-17