[npinsker]

That’s super fascinating, thanks for the comment. I had no idea you could go so far without a guess.

Making use of the Wikipedia article text (and even the titles) seems mostly, but not entirely, against the spirit of the game; the technique used to solve it here is practically identical to vanilla Wordle. It’d be much harder to code, but ideally it would be able to approach the solution with some sort of walk through an embedding, like humans do.

[gjm11]

I think it would be very interesting to try to teach a computer to play the game without having explicit knowledge of what's in all the articles. I'm not sure just what level of ignorance you'd want to enforce: good human players don't have articles memorized but do e.g. have a good idea of the typical structure of Wikipedia articles.

I don't know whether it's feasible to get an LLM to play somehow. LLMs are famously bad at letter-counting, but probably have substantial fractions of Wikipedia kinda-memorized. Perhaps it would be possible to fine-tune an LLM to play Redactle, e.g. by giving it a bunch of training data that's a good human Redactle player's musings on games?