Rodrigo Arias Mallo, maintainer of the Dillo web browser, proposes using asciinema terminal recordings as a way to verify that open source contributions are made by humans rather than AI. The idea is that LLMs struggle to convincingly fake realistic programming session recordings because the training corpus for such recordings is much smaller than for code and patches. Dillo is not yet requiring this, but Arias plans to test the approach further.
Sort: