Generation 1

Darwin 0.1 — Evolution Infrastructure

The first generation establishes evolution itself. Not a model that evolves — an infrastructure that enables any Mind to evolve.

Evolution Infrastructure

✓ Complete

Candidate generation, sandboxing, benchmarking, reviews, promotion, rollback

Mind Runtime

✓ Complete

Integrated system with task execution, memory, skills, tools, routing

Cognitive Compiler

✓ Complete

Goal → Task Graph with optimization and caching

Memory Engine

✓ Complete

10 memory types with full lifecycle management

Skill Engine

✓ Complete

Composition, validation, built-in templates

Tool System

✓ Complete

Capability-based with sandboxed execution

Model Routing

✓ Complete

Policy-driven with privacy gates

Genome Versioning

✓ Complete

Diff, branch, rollback, reproduce, lineage

Benchmark Engine

✓ Complete

MindBench with 20 metrics and provenance

Ollama Runtime

⟳ In Progress

Model discovery, loading, generation

MLX Runtime

⟳ In Progress

Apple Silicon optimized inference

llama.cpp Runtime

⟳ In Progress

GGUF models with quantization

MicroMind Experiment 001

○ Planned

A-K comparison: 0.8B variants vs 2B/4B raw

M0Active

Foundation Stabilization

  • Fix TS strictness issues
  • CI/CD with build order
  • Pre-commit hooks
M1Next

Mind Runtime Hardening

  • E2E task execution
  • Memory persistence
  • Skill/tool composition
M2

Model/Provider Integration

  • Ollama integration
  • MLX for Apple Silicon
  • llama.cpp server
  • Local model benchmarks
M3

MicroMind Experiment 001

  • Run A-K comparison
  • Publish results
  • Analyze efficiency
M4

First Evolution Demo

  • Identify weakness
  • Generate candidates
  • Sandbox + benchmark
  • Human approval → promote

Explore the Full Darwin Specification

Read the complete technical specification, architecture diagrams, and implementation details.

One Command
seai init → seai run
Local First
Runs on your hardware
Evolves
Improves through experience