About

OMB has failed, for 15 years, to produce a complete federal program inventory (FPI). This demo shows a federal program inventory that OpenAI produced in one day.

It’s probably mostly correct. It’s almost certainly not 100% correct. But it’s a place to start. And the shift from blank page to first draft should be enough to push past the inertia.

This FPI has a rational structure with a more complete data framework than the official FPI, because this one is designed to answer the question of: “What does the government actually do, and how does it apply our tax dollars to do that?”

It’s about illuminating Statute → Program → Budget → Outlay. That makes it something that decision-makers can actually use. (No single government resource provides this view today.)

The author hopes that this demo:

1. Sparks a conversation about  how to catalyze action among agency leadership
2. Adds to the conversation about how to leverage LLMs in a not-ridiculous way when it comes to administrative reform
3. Is actually used by OMB as a first draft for agencies to work from, because the data might not be right but the data framework, methodology, and CSV template are objectively much better than the current state

This federal program inventory was developed in a program-first way — using the US Code and FY24 agency performance reports as the primary sources — so that it (probably) accounts for all programs instead of the subset the official FPI pulls from SAMS and USAspending.gov. 

So this demo explains where all of the FY2024 federal budget went, and rolls a lot of smaller expenditures up to meaningful program levels. It also includes financial info for DoD and other tricky departments. And where the public data isn’t sufficiently detailed to break down spending by program, it uses a thoughtful and transparent estimation process to map them. Sources are detailed per row in the CSV download, each individual program page on this demo site, and briefly explained at a high level on the Data Details page.