Summary

Wes McKinney introduces the 'Clanker Constitution,' a set of operating principles for coding agents developed at Kenn through hands-on experience tuning AI behavior across projects. The term 'clankers' deliberately avoids calling them 'agents' to resist over-anthropomorphizing them. The constitution is structured as seven behavioral directives covering request fidelity, autonomous judgment, task completion, work preservation, reality verification, human communication, and durable learning. It is published openly under CC BY 4.0 for others to include in repositories or system prompts. The core thesis is that coding agents require explicit behavioral governance to be reliably productive, and that this governance is itself significant engineering work.

Key Insight

Coding agents require explicit behavioral constitutions to be reliably productive—governing when to act autonomously, when to ask, and how to verify reality is itself serious engineering work that teams must invest in.

Spicy Quotes (click to share)

• 8

  Getting coding agents (we say 'clankers': 'agents' gives them too much credit) to behave in a reasonable manner is a full time job even with the latest frontier models.

• 2

  Treat explicit instructions and constraints as a contract.

• 3

  Scale process to the task. Do not impose specification, planning, or approval ceremony on straightforward work.

• 2

  Pursue the requested outcome until it is verified or genuinely blocked.

• 4

  Never claim success without fresh evidence. Distinguish verified facts, inferences, and unverified assumptions.

• 3

  Test behavior and contracts, not source text, configuration tautologies, or mocked versions of the same logic.

• 2

  Do not create agent-private memories instead of updating shared instructions.

• 3

  Describe pull requests as they exist now, not as a history of discarded approaches.

Tone

pragmatic, prescriptive, opinionated