Hazard Watch
read-only

Documentation

How the world model works

Design, definitions, and a full adversarial audit of our own methods. Everything is version-controlled alongside the code; these link to the source of truth on GitHub.

Start here

HANDBOOK.md

Handbook

The single source of truth: orientation, product & commercial roadmap, engineering design, the accuracy & moat program, and the epics-and-tickets backlog.

Read on GitHub →
GLOSSARY.md

Glossary

Plain definitions of the terms used across the project.

Read on GitHub →

Accuracy & evaluation

EVALUATION.md

Evaluation

One held-out metric per hazard — the key accuracy number for each.

Read on GitHub →
docs/archive/METHODOLOGY_REVIEW.md

Methodology review

A literature-grounded, adversarial audit of every method; 40+ gaps, each graded.

Read on GitHub →

Data & runbooks

DATA.md

Data

The datasets, where each can run, and the reproducibility rules.

Read on GitHub →
docs/ACQUISITION.md

Acquisition

Runbook for pulling the external data (Earth Engine, SAR, OSM) the planned layers need.

Read on GitHub →
docs/AUTOMATION.md

Automation

How the whole pipeline runs free — GitHub Actions + one persistent Postgres, and how the console is served.

Read on GitHub →

Prefer the machine-readable version?

Every number on the map is a static JSON file you can download, with no key and no account.

Data feed →