It's not resistant to adversarial tweets, but it usually collects enough tweets to have an answer in around 1 minute, so it's not too bad to restart if some bad tweets were sampled.
Maybe I should try to use your wordle-tweets dataset to make it work offline as well. :)
I did something similar last week using the Twitter Stream API: https://github.com/basile-henry/twitter-wordle
It's not resistant to adversarial tweets, but it usually collects enough tweets to have an answer in around 1 minute, so it's not too bad to restart if some bad tweets were sampled.
Maybe I should try to use your wordle-tweets dataset to make it work offline as well. :)