A 4B coding agent, post-trained only with RL on synthetic software tasks, reaches 61.5% on SWE-bench Verified and beats models many times larger. No distillation from a frontier teacher. The missing ingredient is not a bigger model. It is tasks calibrated to what the current checkpoint can almost, but not quite, solve.

The pipeline estimates how often the agent already resolves a candidate task, then rewrites the ones that are too easy or too hard until they sit in a learnable band. As the policy improves, the curriculum moves: from heavily scaffolded bug fixes toward shorter statements and more diverse change types. A simple tool harness (read, write, edit, glob, bash) lifts the same 4B base far more than it lifts large models. Failures remain mostly reasoning, not tool use. Effective reward hacking is held at zero by the harness, not by hoping the agent is honest.

Most review processes for agent skills still treat them as documents. They check structure, style, and security. They do not answer the deployment Liftquestion: does this skill actually help a live agent complete real tasks under the same model, sandbox, and grading policy?

ACES (Agentic Continuous Evaluation of Skills) answers that question with paired live trials. For each task the system runs the agent once with the target skill and once without it. Trajectories are normalized, graded on multiple runtime metrics, and compared. The central measure is Skill Lift: the measured improvement the skill produces under fixed conditions.

In 1494, Luca Pacioli described double-entry bookkeeping. Every transaction entered twice. Debits must equal credits. The ledger catches the error.

Five hundred years later, that system remains the decisive advantage for AI-native professional services.

Existing agent protocols (like A2A and MCP) leave lifecycle management, version tracking, and safe evolution underspecified — leading to brittle, monolithic systems.

Autogenesis Protocol (AGP) fixes this by cleanly separating what evolves from how it evolves: