Why adding more sub-agents makes your pipeline worse: reliability comes from deterministic code
Aggressive-Show-281 · reddit · 2026-11-19
Reddit user Aggressive-Show-281 argues most teams misusing agent frameworks chain five autonomous agents for research, critique, coding, and testing — and within three turns error rates compound and token costs explode: agent three hallucinates an API parameter, agent four builds a broken wrapper around it, agent five signs off because its prompt told it to be agreeable.
The core claim: production reliability never comes from letting models talk to each other. Replace agent fluff with deterministic code — standard Python state machines, strict JSON schemas for tool calls, and a single LLM call only where you genuinely need unstructured reasoning. A sober counterpoint to the multi-agent orchestration hype.
More from coding & agent
- Apple opens iWork to external LLMs: Pages, Numbers, Keynote gain MCP support — dsp_ · 2026-11-19
- Anthropic engineer shares free 30-minute Claude Code SDK + GitHub Actions automation walkthrough — aftahi_ai · 2026-11-19
- Snowflake Labs open-sources Cortex AI 101 notebook calling OpenAI models via SQL — adnan_hashmi · 2026-11-19
- Agents spent years hillclimbmaxxing while the human at the keyboard became an afterthought — tokenbender · 2026-11-19
- Google's GUARD paper: frozen base model, self-editing agent harness gains +8.89 points — omarsar0 · 2026-11-19
- Microsoft open-sources openclaw-dev: one-command hosted OpenClaw agent on Azure with sandboxed tool execution — adnan_hashmi · 2026-11-19