What would it take to trust coding agents running overnight?
Low_Cable2610 · reddit · 2026-11-22
A developer juggling Claude Code, Codex, and Cursor on one project describes acting as the middleman: writing prompts for each, requesting reports, and pasting them between tools. He's building a small desktop app that takes one goal, dispatches it to the agents he already pays for, and keeps a shared memory across them.
The open question is unattended overnight runs: hand it a task before bed, wake up to a receipt of what changed, what it cost, and which tests passed or failed. He admits he doesn't yet trust that — his own or anyone else's.
He asks the community: what would you need to see in the morning to trust unattended agent work, whether stuck agents should halt and wait or keep going with a flag, and what broke the first time you tried overnight runs.
More from coding & agent
- Sharing the Composer cache in GitHub Actions made PHPUnit integration tests 31% faster — DanielLockyer · 2026-11-22
- Google's Mirror Dimension: synthetic web for safe browser-agent testing — shashib · 2026-11-22
- Exposing Google/YouTube trending searches to agents via an MCP tool — Substantial-Tour-581 · 2026-11-22
- What hard stops do you put on your AI agents? A practical red-line list — Virtual_Arachnid7916 · 2026-11-22
- Stardock CEO shows Clairvoyance: hierarchies of 71 AI 'staff' agents building Sins of a Solar Empire II — draginol · 2026-11-22
- FieldCraft: pure-Rust open-source CAE tool bundles FEA, heat and fluids with MCP support for AI agents — Promptmethus · 2026-11-22