← 返回AI教程
🌐 其他

Claude Code Runs the Real Ponytail. Cursor and 11 Others Settle for 2,593 Bytes.

来源:Towards AI · 发布于 2026-08-19 16:00:28
Claude Code Run
Last Updated on August 19, 2026 by Editorial Team Author(s): Chew Loong Nian – AI ENGINEER Originally published on Towards AI. Claude Code Runs the Real Ponytail. Cursor and 11 Others Settle for 2,593 Bytes. Ponytail’s own portability doc lists 22 coding agents. I parsed it and counted: only 9 of them get an adapter the agent actually executes. The other 13 get no runnable adapter at all, and twelve of those get nothing but the same rule body — the 2,593-byte AGENTS.md, or a byte-identical copy of it filed under their own name. That is 39% of the real ruleset, missing every one of its five behaviors. The rest of the article explains how the author computed which hosts truly execute Ponytail’s adapters versus merely read a copied rules file, including the Python script used to classify hosts and the resulting 9 EXEC vs 13 TEXT split (with caveats where installs or manifests blur the edges). It then details what Ponytail is in practice—a decision ladder with safety/guard-rail requirements—and why the “text tier” loses major behavioral controls like intensity levels, off switching, persistence, output contracts, and subagent coverage (notably affecting Cursor and others). The author also verifies that the distributed copies don’t drift by running a CI check and re-hashing rule files, analyzes how Ponytail’s large repository is mostly benchmarks and tests rather than pure adapter boilerplate, and highlights mismatches between what the badge claims and what’s actually documented in agent portability lists. Finally, it compares Ponytail’s adapter “tier” model to how different agent harnesses handle plugins, and concludes with a practical “which should you use?” breakdown, arguing that star counts are misleading and that distributing real behavior across 2026 agent ecosystems requires more than sharing a markdown file named AGENTS.md. Read the full blog for free on Medium. Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date

Last Updated on August 19, 2026 by Editorial Team Author(s): Chew Loong Nian – AI ENGINEER Originally published on Towards AI. Claude Code Runs the Real Ponytail. Cursor and 11 Others Settle for 2,593 Bytes. Ponytail’s own portability doc lists 22 coding agents. I parsed it and counted: only 9 of them get an adapter the agent actually executes. The other 13 get no runnable adapter at all, and twelve of those get nothing but the same rule body — the 2,593-byte AGENTS.md, or a byte-identical copy of it filed under their own name. That is 39% of the real ruleset, missing every one of its five behaviors. The rest of the article explains how the author computed which hosts truly execute Ponytail’s adapters versus merely read a copied rules file, including the Python script used to classify hosts and the resulting 9 EXEC vs 13 TEXT split (with caveats where installs or manifests blur the edges). It then details what Ponytail is in practice—a decision ladder with safety/guard-rail requirements—and why the “text tier” loses major behavioral controls like intensity levels, off switching, persistence, output contracts, and subagent coverage (notably affecting Cursor and others). The author also verifies that the distributed copies don’t drift by running a CI check and re-hashing rule files, analyzes how Ponytail’s large repository is mostly benchmarks and tests rather than pure adapter boilerplate, and highlights mismatches between what the badge claims and what’s actually documented in agent portability lists. Finally, it compares Ponytail’s adapter “tier” model to how different agent harnesses handle plugins, and concludes with a practical “which should you use?” breakdown, arguing that star counts are misleading and that distributing real behavior across 2026 agent ecosystems requires more than sharing a markdown file named AGENTS.md. Read the full blog for free on Medium. Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor. Published via Towards AI