Research project investigating intelligence pricing trends and dynamics.
You primarily work in three places:
plans/todo.md— the master to-do list. Add items, reprioritize, check things off. This is where you steer the project.plans/active/— execution plans for complex tasks that need their own scope and tracking. Most to-do items don't need one.
Write and revise paper sections. Each section is a standalone markdown file. The master document drafts/main.md assembles them in order.
To preview the current draft as HTML:
python3 drafts/render.py
open drafts/draft-$(date +%Y-%m-%d).html
Define what must be true for the paper to hold up under review.
tests/master.test.md— quality criteria that apply to every section (clarity, flow, claim-evidence alignment).tests/<section>.test.md— reviewer simulation for each section. Enumerate the critiques a reviewer would raise and track whether the draft addresses them.tests/model-paper.test.md— benchmark against an accepted high-quality paper. What standard does it set, and does our draft meet it?
Open progress.md — latest entry is at the top. This is the audit trail of everything that has been done.
CLAUDE.md agent instructions (philosophy, conventions)
setup.md agent setup guide (bootstrapping, session start)
README.md this file — human instructions
progress.md reverse-chronological progress log
plans/
todo.md master to-do list
active/ execution plans for complex tasks
completed/ finished plans
tech-debt-tracker.md known shortcuts and deferred work
drafts/
main.md assembles sections via :(sections/file.md)
sections/ individual section markdown files
render.py markdown → HTML renderer
draft-YYYY-MM-DD.html dated HTML snapshots
tests/
master.test.md cross-section quality tests
model-paper.test.md model paper benchmark
*.test.md per-section reviewer simulation
code/ scripts, pipelines, analysis code
data/ datasets and derived outputs
runs/ run-specific logs, checkpoints, artifacts