Implementer–reviewer: running agents that review other agents for better output
Al_Grigor · x · 2026-10-21
Alexey Grigorev shares his agentic engineering practice: one agent implements, another validates against shared rules (design system, Clean Code), iterating until the reviewer accepts. Even frontier models produce pages of slop, so he uses agents to control agents — trading 20 minutes for a few hours but getting better quality. A survey cited in his article found 59% of respondents whose workflow was 'fundamentally' changed by AI use another AI/agent to critique outputs, vs 54% in the 'significantly' changed group. The article also covers giving agents coding/writing rules, organizing agents into hierarchies, and software factories where humans focus on outcomes.
Related event: Agentic Engineering: Humans Write Specs, Agents Code and Review(2 posts)→
More from coding & agent
- Microsoft's Guardian Agents treat AI agents as insider risks to cap runaway permissions — emmanuelvivier · 2026-10-21
- Dev builds attest: position-aware prompt-injection scanner for MCP tool descriptions — mrpkk · 2026-10-21
- S-Agent 2.0 turns coding agents into spatial reasoners, hitting 66.5 on MMSI-Bench — liuziwei7 · 2026-10-21
- LLM trading agent skips 380 of 392 calls — scoring refusals shows skipped events moved 2x random — Tasty-Mammoth5321 · 2026-10-21
- Seven single-job Grok bots beat one do-everything content agent, creator says — lucasbennett_1 · 2026-10-21
- Castov: an open-source zero-telemetry MCP server and DevSecOps CLI — benidira · 2026-10-21