About GDPQuery.ai
GDPQuery is a research tool for United States macroeconomic data. Ask a question in plain English and it returns the official series as a precise chart alongside a sourced analyst brief — drawn from FRED, the Bureau of Economic Analysis, the Bureau of Labor Statistics, and the Census Bureau.
Founder
Wade Petty — Founder, Cedar Mountain Systems
I’m Wade Petty. For fourteen years I helped build the systems behind the United States’ official economic statistics, most of that time at the Bureau of Economic Analysis. I started as a developer building pieces of those systems, was promoted to lead them, and now work independently as Cedar Mountain Systems LLC — GDPQuery is my first product.
At BEA I led a team of eleven — data scientists, economists, and technical staff — responsible for the SQL system and front end behind national GDP: it aggregates the detailed estimates produced by the agency’s economists into the published figures and supports the internal review each release goes through. We were also responsible for the GDP-by-Industry and Input-Output accounts. I led the project that rebuilt the GDP-by-Industry system, cutting its production time by more than half and retiring a stack of legacy desktop processes in favor of centralized, inspectable SQL — work that let GDP-by-Industry publish alongside the final quarterly GDP estimate and earned our branch chiefs a Department of Commerce Gold Medal. I also served on the committee that recommended Python as BEA’s standard processing language, and co-chaired the team that put it into practice.
Why I built GDPQuery
These are numbers the public paid to produce, and they should be easy to reach — especially now, when it’s tempting to ask an AI model for a figure it will give you confidently and get wrong or out of date. GDPQuery takes you to the actual, current official number, so a journalist or a citizen can check a claim against the real data. It draws only on figures after they’re publicly released — the same numbers, on the same schedule, available to everyone.
Looking up the number and charting it is free for anyone — that’s the point. The paid tiers add on-demand analysis, exports, and the tools to build on the data.
How it works
I designed and built GDPQuery end to end: a Next.js front end, a FastAPI backend, PostgreSQL on Google Cloud, and a dual-agent LLM pipeline on Google’s Gemini models — one agent retrieves data and writes SQL, another synthesizes the analysis. Where language models are unreliable — arithmetic on the numbers — deterministic code does the math, and every model-written query is validated against a least-privilege database role before it runs. I treat model output as untrusted and bound its cost by design.
Sysop
To ship at that volume safely, I built the review into the process. Sysop is an open-source framework that keeps AI-assisted development honest with deterministic gates and adversarial plan review. It runs across my work and is public at getsysop.com (source at github.com/getsysop/sysop).