checkgov: Macrocast

Our previous CheckGov experiment was ambitious, but as the trial continues, we wanted to push further in a different direction.

Last time, we focused on unstructured data and the natural language processing required to make sense of it. This time, we saw an opportunity in the opposite challenge: highly structured data, grounded in mathematical methods of predictive modeling.

That led us to begin building Macrocast, a system designed to serve as a durable foundation for a wide range of financial and economic forecasting applications over time. Macrocast already draws extensively from link@tadalabz/shanti and currently supports 849 distinct fact data types sourced from FRED, BLS, BEA, Census, and Treasury.

It is still early, but the foundation is taking shape. Our next milestone is finalizing that first data collection sweep and turning that data infrastructure into our next user-facing CheckGov asset, our own GDPNow-style forecasting model, a model that ingests newly released economic data and updates its estimate of current GDP growth in near real time.

We are not there yet, it's still just one part of a yet to be successful experiment in the larger trial, but the progress so far felt worth sharing.

See you soon.