I built a dev tool that 'evolves' code with AI — REAP

Published: (March 17, 2026 at 10:57 PM EDT)
2 min read
Source: Dev.to

Source: Dev.to

The Problem

If you’ve been building with AI agents (like Claude Code), you’ve probably encountered these problems:

  • Context loss – Starting a new session wipes the context, forcing you to cling to long sessions just to avoid losing everything the AI has learned.
  • Stale documentation – Persisting knowledge in READMEs and CLAUDE.md files quickly becomes outdated as the project evolves.
  • AI going rogue – Sometimes the AI ignores carefully crafted docs and does its own thing anyway.

We’re all stuck at the same bottleneck – the context window just isn’t enough for long‑running projects.

Existing tools like spec‑kit and superpower are decent for one‑off feature work, but they don’t fit sustained, long‑term development.

What I Built

I built REAP (Recursive Evolutionary Autonomous Pipeline) – an open‑source CLI tool inspired by generational evolution in biology.

The idea: AI and humans evolve software across generations.

Genome (Design & Knowledge)
  → Evolution (Generational Progress)
    → Civilization (Source Code)

How It Works

Genome

Your project’s design knowledge is managed as a Genome – architecture decisions, business rules, conventions, and constraints.

.reap/genome/
├── principles.md      # Architecture principles
├── domain/            # Business rules
├── conventions.md     # Development conventions
└── constraints.md     # Technical constraints

Life Cycle

Each generation follows a five‑stage lifecycle:

Objective → Planning → Implementation → Validation → Completion
  • Objective – Define goal, requirements, and acceptance criteria.
  • Planning – Break down tasks, choose approach.
  • Implementation – Build with AI + human collaboration.
  • Validation – Run tests, verify completion.
  • Completion – Retrospective, apply Genome changes, archive.

Evolution

  • When a generation completes, it is archived in the lineage, and the next generation picks up new goals.
  • Lessons learned within a generation are folded back into the Genome.
  • Through this iterative pipeline, your source code (the Civilization) keeps evolving.

Quick Start

# Install
npm install -g @c-d-cc/reap

# Initialize
reap init my-project

# Run a full generation in Claude Code
claude
> /reap.evolve "Implement user authentication"

/reap.evolve runs the entire generation lifecycle – from Objective through Completion – interactively with you.

  • GitHub:
  • Docs:

MIT licensed. Contributions and feedback are welcome!

0 views
Back to Blog

Related posts

Read more »

You're not prompting it wrong.

Background I was listening to Grady Booch on The Third Golden Age of Software Engineering episode of The Pragmatic Engineer. During the episode he mentioned a...