An 83-Page Teardown of 11 Coding Agents: Three Findings, Two Hands-On Checks, One Expired Constant
AI disclosure: this article is AI-generated — written by my agent toolchain under my direction. I set the questions, ran the hands-on checks, and verified every claim; the judgments are mine. Wavestone AI Lab has published an 83-page source-code study of eleven production coding agents: Claude Code, Codex CLI, Gemini CLI, Mistral Vibe, OpenHands, Aider, Mini-SWE-Agent, Hermes, Pi, OpenCode, and…
An 83-page study from Wavestone AI Lab analyzes code from 11 coding agents, including Claude Code, Gemini CLI, and Hermes. The researchers found that none of the systems use a general-purpose agentic framework or embed code retrieval into their context loop. Nine of the 11 systems adopted a skills/plugin system, with one example of a skill authored by the agent itself.
The study also found that supply-chain governance is becoming more important as architectures converge and older mechanisms decay. The researchers verified two of the paper's claims themselves and found that a constant used by Claude Code had been renamed in the shipping binary. They also used Hermes, one of the systems analyzed, to calibrate the findings presented in the paper.
The consultancy likely published the study to build credit and credibility in the AI coding agent market, which includes product offerings, client case studies, hiring, financials, and partnerships.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
