This solver treats ARC-AGI-2 as a search problem inspired by Grover's search algorithm. The approach uses program synthesis with multiple iterations where each iteration generates candidate solutions, grades them against training examples (similar to an oracle), and rotates the context window to place the best attempts last (most recent in context). Preliminary results on 60 sample problems from the ARC-AGI-2 evaluation set show 38% success rate.
Hello HN! C++ interop is a project I've been working on for a year and a half now, so you have no idea how cool it is to see this on the front page. Really made my day. Anyway, there are some really great questions in this thread (and some slight mis-understandings), if you want your questions answered, feel free to post them as a reply to the announcement on the Swift forums (linked above). I'd be happy to answer them :)
(To keep things organized, it would be great to have all the Q&A happen in one place, that's why I'm asking you post it in the forums. And remember, every question is valid, so don't be shy!)