Replication package for: Measuring AI diffusion in Latin American firms: evidence from US securities filings

Replication package for the article “Measuring AI diffusion in Latin American firms: evidence from US securities filings” (Carlos Miguel Yalta Vargas and Kangjuan Lv). It contains the frozen source pulls (SEC EDGAR annual reports and XBRL data, exchange rates and prices), the constructed firm panel, and the Python code that rebuilds every table, figure and number in the article from these inputs. To reproduce, run python3.11 -m pip install -r panel_c/requirements.txt, then python3.11 panel_c/run_panel_c.py --to manuscript_values (about 50 minutes). To check the reproduction afterwards, run python3.11 -m pytest panel_c/tests -q -m "not record". README.md describes the pipeline, GUARDRAILS.md the checks, and MANIFEST_SHA256.txt gives the checksum of every file. The code is MIT-licensed and the constructed data CC BY 4.0 (LICENSE.md).