Build And Run The Atlas

Engine smoke run

The engine favors clarity over speed. It stores state as JSON so history can be read without special tools. It writes one folder per generation with verdict and lesson files. It caps patch size to keep review human scale.

Create a Python virtual environment and install the minimal requirements. Run the loop in dry mode for two generations. Read the diary and verdict files. Confirm that rejected tries carry clear reasons. Keep the first real experiment narrow with a fast test.

Docs app setup

The docs app uses Next.js App Router with a fixed left sidebar on desktop and a calm top bar on small screens. The palette rests on slate, emerald, teal, amber, cyan, and warm neutrals. Type scale is steady and spacing is generous.

Move into the docs folder and install dependencies with a clean cache policy. Run the development server and open the local address. Walk the sidebar from Start through Safety to Roadmap. Confirm that search filters by title and summary without a server round trip.

First safe experiment

Pick one narrow task with a fast test. Add one fixture with a hidden check. Run one generation with a small change budget. Keep the patch only when the full suite passes. Archive the rest with reasons and share the diary entry with a reviewer.

When a generation succeeds, raise difficulty by one notch. When a generation fails twice on the same pattern, step difficulty down and grow the lesson archive. Degenerate tasks with no hidden check are rejected at creation time.

Repository layout

Engine dry run

py engine/rsi_loop.py --state engine_state --generation 1 --files worker/planner.py --changes 1 --passed true --gain 0.2 --cost -0.1
py engine/rsi_loop.py --state engine_state --generation 2 --files judge/judge.py --changes 1 --passed true --gain 0.4 --cost 0.0
type engine_state\history.json

Docs app commands

cd docs
npm install --package-lock=false
npm run dev
npm run build

Continue reading