Dark code
The DeepSeek V4.1 technical report has this interesting tidbit:
Coding agent training environments are built from … public GitHub repositories that meet a star-count threshold. Environment construction is carried out collaboratively by multiple specialized agents. First, an agent determines whether the project can be built and fully run inside a container and whether it can be automatically verified; if so, it selects a specific turn or commit as the task starting point, designs several sufficiently complex implementation directions, and produces concrete evaluation points, including both fail-to-pass and pass-to-pass points, along with a construction report, fetching external resources from the web as needed. Next, a separate agent sets up dependencies, the initial working directory, test code, and task descriptions in an isolated container, performs self-testing, removes any traces that could leak the task solution, and packages the environment as a new image layer. Then, multiple distinct agents attempt the task, and an independent quality-inspection agent reviews the environment together with the solving agents’ trajectories, …
So essentially as part of model training the AI models solve random problems in github repositories. If you’re working on a public github project with enough stars, and looking at a problem, some AI agent might have already solved it somewhere in the dark. Maybe the solution was bad, or maybe it was better than yours.