Source-code study of 11 coding agents finds every harness shares the same 7 parts
blaizedsouza · x · 2026-11-04
Engineers opened the source of 11 real coding agents (Claude Code, Codex CLI, Gemini CLI, OpenHands, Aider, Hermes) and concluded: agent = model + harness, with every system built from the same 7 components.
The attached arXiv paper (Wavestone AI Lab / Inclusive Brains) details:
- The 7 canonical parts: agent loop, LLM integration, tools & actions, memory & context, safety & permissions, orchestration (sub-agents), extensibility.
- 13 cross-cutting observations and 29 recurring design patterns; SKILL.md and agent-led MCP adoption at 9/11 and 8/11; OpenHands runs Claude Code, Codex, or Gemini CLI as interchangeable backends.
- The quoted article describes a hands-on setup: an AI employee handles the night shift under a strict "do not merge or change anything on your own" rule — it reads, compares, diagnoses and drafts PRs, the human decides.
More from coding & agent
- ThunderAgent merged into LLM-D: 2x agent serving throughput, TTFT from ~50s to 2-4s — simran_s_arora · 2026-11-04
- Agents pass the test but fail the job: the requirement gap and the model gap — bigdata · 2026-11-04
- Does that new prompting trick actually work, or just look better? — jimmykoppel · 2026-11-04
- Vercel launches Security Review in public beta, free until Nov 10 — cramforce · 2026-11-04
- QuickProof: formal verification for C with specs written as plain C tests — thomasahle · 2026-11-04
- Project Moah: Open-Source Memory Linker That Gives AI Agents a Brain — starcount_games · 2026-11-04