%%{init: {"theme": "base", "themeVariables": {"primaryColor": "#f8fbf7", "primaryTextColor": "#1f2d29", "primaryBorderColor": "#7a9b76", "lineColor": "#7a9b76", "tertiaryColor": "#ffffff", "fontFamily": "Poppins, sans-serif"}}}%%
flowchart LR
sources["Public web sources"] --> skills["Two Claude Code skills"]
skills --> engine["Shared radar engine"]
engine --> review["Source review and scoring"]
review --> validators["Deterministic validators"]
validators --> library["Structured library"]
validators --> queues["Human review queues"]
library --> reports["Daily reports"]
classDef stage fill:#f8fbf7,stroke:#7a9b76,color:#1f2d29;
classDef output fill:#ffffff,stroke:#2f5d50,color:#1f2d29;
class sources,skills,engine,review,validators stage;
class library,queues,reports output;
Harry Does AI
AI systems for social science and policy
Public AI workbench
I am an economist building AI systems for social science and policy.
My goal is to help social scientists and policymakers navigate research, evaluate evidence, and make better decisions.
Projects
AI Radar
The AI Radar is a self-running research system that tracks AI developments that matter to social science and AI implementation.
Every morning, two Claude Code skills scan public sources and save the most useful findings as structured library entries. The daily reports work like a newsletter tailored to specific research preferences, and the library keeps the selected work as a growing source of knowledge.
Running daily since July 2026, with 130+ structured entries and 30+ daily reports in the public repository.
Judgment
Domain profiles define what counts as useful for social science and AI engineering.
Guardrails
Python validators protect reports, paths, queues, metadata, and append-only history.
Archive
Reports, library entries, rejections, source proposals, runbooks, and incidents are public.
How it works
AI scans the sources and makes the first review. Before anything enters the library, the code checks that it follows the rules. If a source is unclear, the system sends it to human review.
Public skills
AI Engineering Radar
For methods and practices used to build reliable AI systems, research agents, and policy products.
Social Science Radar
For AI developments that matter to economists, social scientists, and policy researchers.
Open skill