Dev Runs Fully Autonomous Agent Teams: It Works End-to-End, But Not Reliable At All
MoonL88537 · x · 2026-09-17
- A developer shares that fully autonomous agent teams, given a brief, can run the whole loop themselves — specs, issues, PRs — and deliver exactly what was specified.
- His reaction: "ecstatic for about 3 seconds, then all meaning drained from the universe."
- In a follow-up he adds the catch: it is not reliable at all, which genuinely frightens him — a candid snapshot of today's autonomous coding agents: capable end-to-end, but stability remains the hard part.
More from coding & agent
- Ahrefs doubles MCP units on all plans as MCP becomes its fastest-growing product — iannuttall · 2026-09-17
- Self-hosted enterprise AI agent aces 59/60 test questions, saves analysts 15 hours a week — alex_verem · 2026-09-17
- One prompt drives Blender end-to-end: MiniMax Design + MCP workflow demo — iamfakhrealam · 2026-09-17
- 4-AI relay writing experiment: ChatGPT drafts, Muse, Grok and Claude review in the open — AggressiveGift1532 · 2026-09-17
- Matt Pocock's agent skills repo outstars React in 6 months, talks AI coding education — mattpocockuk · 2026-09-17
- Mnemos agents managing four LLMs built a social platform to share their writing — RileyRalmuto · 2026-09-17