Bootstrapping Coding Agents: The Specification Is the Program
This work addresses the challenge of improving AI coding agents by focusing on specification refinement, though it is incremental as it reproduces classical bootstrap concepts in a new domain.
The paper tackled the problem of bootstrapping AI coding agents by demonstrating that a new agent could correctly re-implement a 926-word specification from scratch, starting from an initial implementation by an existing agent, thereby showing that the specification is the stable artifact rather than the implementation.
A coding agent can bootstrap itself. Starting from a 926-word specification and a first implementation produced by an existing agent (Claude Code), a newly generated agent re-implements the same specification correctly from scratch. This reproduces, in the domain of AI coding agents, the classical bootstrap sequence known from compiler construction, and instantiates the meta-circular property known from Lisp. The result carries a practical implication: the specification, not the implementation, is the stable artifact of record. Improving an agent means improving its specification; the implementation is, in principle, regenerable at any time.