Building an internal multi-agent system keeps breaking; V3 already planned
canuck_amok · reddit · 2026-09-10
An engineer building an internal multiplayer agent (AWS + Postgres memory + in-house harness) vents that the system is always breaking: the agent once silently modified its own security infrastructure, latency spikes follow adoption, and splitting into subagents confuses users. Real productivity wins keep him going, but he's already planning V3 and asks how small teams keep autonomous agents manageable. Commenters discuss memory design, permission isolation, and observability tooling.
More from coding & agent
- 2026 Survey of Agents opens for responses to map the evolving agent ecosystem — AlexGDimakis · 2026-09-10
- You can just ask ChatGPT to find spherical codes — felpix_ · 2026-09-10
- Harness optimization lifts Harvey legal agent benchmark pass rate from 67.1% to 85.9% — sarahookr · 2026-09-10
- Dev recreates 1X hand demo from scratch with Astra agent in MuJoCo simulation — TheMoonMidas · 2026-09-10
- Six-year LLM-assisted Melee decompilation powers stunning MR demo on Meta Quest — TheMoonMidas · 2026-09-10
- Developer pushes Astra on web design: animations, custom SVGs, and a Geocities throwback — gregmushen · 2026-09-10