Skip to content

aintoniodev/genetic-programming-pipelines

v1.0.0MIT

Genetic programming (GP / programación genética) applied to neural-network and deep-learning pipelines: evolutionary program-space search (code, trees, configs, architectures, augmentations, prompts, agents) with fitness-ladder design, LLM operators, champion selection, diversity pressure, and objective audits. Activates only on genetic-programming topics, not generic autoresearch.

GitHub

genetic-programming-pipelines

social preview

Genetic programming (GP) applied to neural-network and deep-learning pipelines: evolutionary search in program space (architectures, augmentations, configs, prompts, agents) with LLM variation operators, fitness ladders, deterministic gates, champion selection, diversity pressure and objective audits.

  • Activates only on genetic-programming topics (ES: programación genética), not generic autoresearch, HPO or NAS without a GP framing.
  • Evidence base: 41-source scientific synthesis (2026-08-09) + field validation on a live benchmark (2026-08-10: 18 CNN-MNIST candidates, 3 generations, RTX 2060, rank-inversion and tier-blind-spot findings logged in skills/genetic-programming-pipelines/references/evidence.md).
  • Plugin format: Agent Plugin 1.0.0 (agent-plugins.org) — plugin.json + skills/ directory.

Install

Point your agent at plugin.json (Agent Plugins compatible client), or copy skills/genetic-programming-pipelines/ into your agent's skills folder (e.g. ~/.pi/agent/skills/ for pi, ~/.agents/skills/ for Claude Code).

Use

Ask your agent for genetic-programming-based optimization of a DL/ML pipeline. The skill guides the loop: search-space framing → fitness ladder (cheap surrogates + full fidelity) → diverse seeding → LLM operators with deterministic gates → champion selection → objective audit → lineage logging.