setup.md

Project Setup — For AI Agents

This file teaches agents how to bootstrap and orient themselves in the IntelligencePriceIndex project.

First-time setup

If the project directories don't exist yet, create them:

mkdir -p drafts/sections code data runs plans/active plans/completed tests

Ensure these files exist:

• drafts/main.md — master document assembling sections via :(sections/file.md)
• drafts/render.py — markdown-to-HTML renderer (zero external dependencies)
• plans/todo.md — master to-do list
• plans/tech-debt-tracker.md — technical debt log
• tests/master.test.md — cross-section quality tests
• tests/model-paper.test.md — model paper benchmark tests
• progress.md — reverse-chronological progress log

For each section listed in drafts/main.md, ensure a corresponding:

• drafts/sections/<section>.md — the draft content
• tests/<section>.test.md — the reviewer simulation tests

Session start checklist

At the start of each session, read these files to orient:

1. CLAUDE.md — philosophy and operating instructions.
2. progress.md — what has been done (latest entry at top).
3. plans/todo.md — what needs to be done next.
4. plans/active/ — any in-progress execution plans.
5. plans/tech-debt-tracker.md — known debt and deferred work.
6. tests/master.test.md — cross-cutting quality status.

Then check git status and git log --oneline -5 for recent changes.

Project structure

CLAUDE.md              agent philosophy and instructions
setup.md               this file — agent setup guide
README.md              human-facing project overview
progress.md            reverse-chronological progress log
plans/
  todo.md              master to-do list (what to work on next)
  active/              execution plans for complex tasks
  completed/           finished plans (moved here when done)
  tech-debt-tracker.md known shortcuts and deferred work
drafts/
  main.md              master document — assembles sections via :(sections/file.md)
  sections/            individual section markdown files
  render.py            script to assemble and render HTML
  draft-YYYY-MM-DD.html  rendered snapshots (dated)
tests/
  master.test.md       cross-section tests (applies to all sections)
  model-paper.test.md  model paper benchmark (separate from section tests)
  *.test.md            per-section reviewer simulation tests
code/                  scripts, pipelines, analysis code
data/                  datasets and derived outputs
runs/                  run-specific logs, checkpoints, temporary artifacts

Paper rendering

Run from project root:

python3 drafts/render.py

This produces drafts/draft-YYYY-MM-DD.html. Re-render after every substantive draft change.

Adding a new draft section

1. Create drafts/sections/<section>.md.
2. Add :(sections/<section>.md) to drafts/main.md in the correct position.
3. Create tests/<section>.test.md with the reviewer simulation template.
4. Re-render the draft.