Hacker News
Two-Thirds of One F-Droid Update Batch Looks LLM-Generated in a Hand Audit of 102 Apps
A September 15 post examined all 102 apps in F-Droid's September 12 update batch and classified 67 (65.7%) as mostly AI-generated, 24 (23.5%) as showing no signs of AI, and 12 (11.8%) as unclear, based on repository aesthetics and commit patterns. The author calls the method superficial, later recategorized about 9-10 apps toward human-coded, and explicitly says LLM usage does not by itself make a project slop. Treat the exact percentage as soft, but the direction matters: this is the first attempt at putting a number on AI authorship in a major open-source app repository.
↳ Follow the thread