sc
wildcatter turns a natural-language question into a standalone Python script that answers it.
Core trick
LLM writes a standalone Python scraper per question, saves it as a rig, and regenerates the code only when a re-run fails.
Description
wildcatter turns a natural-language question into a standalone Python script
that answers it.
You ask something like "revenue per comma in the CEO letter for every Fortune
50 company" — an LLM decomposes it into a plan, writes a Python script that
fetches real data (SEC filings, FRED, Wikipedia, Yahoo Finance, web
scraping), runs it, validates the output with auto-generated tests, and saves
the result as a reusable "rig."
Rigs re-run without the LLM. If something breaks (API changed, site
restructured), the system calls the LLM to fix the code automatically. You
can also set up triggers to re-run on a schedule or when a condition fires.
The output is a structured table you can export directly to excel or via MCP
The linked repo returns 404, so it was deleted or made private.