🌱 Darwin Lab (Refactored)

Meta-evolutionary scientific cognition system — evolving agents, symbolic theories, motifs, and experimental protocols.

---

🔧 Core Formalism

Agent Loop

1. Generate Theory: Symbolic structure with motifs.
2. Evaluate Theory: Against tasks or simulated environments.
3. Attribution: Score subcomponents via TreeLSTM.
4. Mutation: Modify structure based on attribution.
5. Bayesian Optimization: Update strategy weights.
6. Cultural Evolution: Retain, recombine, compress.

---

Key Abstractions

Concept	Role
Agent	Evolves symbolic theories, scored by utility + novelty
Motif	Reusable building block for theories (symbolic, cognitive, neural)
StimLang	Linearized protocol used for experiment/task delivery
HEG	Hierarchical Embedded Graph for reasoning, curriculum, compression
MetaAgent	Synergetic controller — tracks entropy, triggers bifurcations
TreeLSTM	Learns attribution over tree-structured theories

---

📂 Modules

• agents/ → Theory generators, interpreters, and learning models
• evolution/ → Core loop, synergetics meta-layer
• motifs/ → Memory, abstraction, perturbation scoring
• biobridge/ → Kaya interface (astrocyte and neuron feedback)
• tasks/ → Task definitions and curriculum engine
• dashboard/ → Live strategy & attribution visualization
• metrics/ → Theory complexity, compression scoring

---

🔮 Vision

Darwin is a recursive theory engine. It learns how to mutate, how to attribute, and how to structure symbolic knowledge via evolution, optimization, and biological feedback.