Evolve

An open-source LLM-driven genetic programming framework that (kinda) recreates DeepMind's AlphaEvolve. It uses GPT-4.1 to evolve better algorithms, starting from naive baselines—currently focused on matrix multiplication.

No np.dot, no @. Just raw Python loops, strategic mutations, and heartbreak.

Features

• 🧬 Genetic Programming Core
  Tournament selection + mutation-only evolution + Pareto optimization (speed + accuracy).

• 🧠 LLM-based Mutation Engine
  Uses GPT-4.1 to mutate Python code using meaningful, pre-curated strategies.

• 🧪 Multi-Objective Evaluation
  Each candidate is scored by:

  • Runtime performance (in nanoseconds)
  • Accuracy (relative error vs NumPy reference)

• 🔥 Mutation Strategies
  Over 40 real strategies like loop unrolling, loop tiling, Strassen’s algo, blocking, etc. No cosmetic garbage.

• ❌ No Cheating Allowed
  Enforces no use of np.dot, np.matmul, or @.

• 🧱 Modular + Hackable
  Clean and extensible. Add your own strategies, objectives, domains.

---

Installation

git clone https://github.com/think-a-tron/evolve.git
cd evolve
uv sync

Requirements:

• Python 3.10+
• openai, numpy

Set your OpenAI API key in your environment:

export OPENAI_API_KEY=sk-...

---

Usage

Basic run:

uv run main.py --gen 20 --pop 50

• --gen: number of generations to evolve
• --pop: initial population size

At the end, you'll get:

• Final Pareto front (best tradeoffs of speed and accuracy)
• Source code of each front-runner
• The best candidate that’s both fast and accurate

---

How It Works

1. Seed: Starts with a naive matrix multiplication function.

2. Mutation: LLM mutates it using a randomly chosen strategy from a curated list.

3. Evaluation:

   • Runs the code on matrices of size 128x128 and 256x256.
   • Measures runtime and computes error vs NumPy.

4. Selection:

   • Builds Pareto front (minimizing both runtime and error).
   • Uses NSGA-style crowding distance to preserve diversity.

5. Evolution:

   • Surviving candidates are mutated again.
   • Occasional random immigrants added.

6. Repeat.

---

Example Output

Gen 15: best time 2340000 ns, err 1.2e-6

--- Final Pareto Front Solutions ---

Solution 1 (Key: 2d6885552d05 | Time: 2340000 ns | Error: 1.2e-6):

def matmul(a: np.ndarray, b: np.ndarray) -> np.ndarray:
    ...

---

Notes

• Sometimes GPT mutates nothing useful. It’s an LLM, not a genius.
• Reward hacking is real. Without constraints, it will try to “optimize” by doing nothing.
• You can extend this to other domains—FFT, convolution, sorting—if you're brave enough.

---

License

MIT. Use, modify, break, evolve.

---

Contributing

Fork it, mess with it, roast it. PRs welcome if they don't contain a @ b.

---

Credits

Inspired by DeepMind’s AlphaEvolve Built by someone who ran out of API credits multiple times.