Post

In 2026 spelling mistakes are an indicator for human authorship (until genAI starts adding that, too), so no rush. 👍

Are these stored as a bloom filter kind of data structure? (cuckoo filter more likely if you can also remove, but maybe we don't have to?) It'd be great to standardize a few sizes to choose from and let people ship a blob generated once instead of a json with all the words. Tiny fraction of size.

I used a dirty hack to make OS agnostic app that added spellcheck in multiple languages to textfields on the same page in our webapp in a human translator aid tool in 2009. Still impossible out of the box in today's browsers AFAICT There's definitely room for improvement. And a shrinking audience🙈