Elektrine lite

← Feed

@KayOhtie@blimps.xyz

Post #1225395

2026-04-05 15:37 UTC

@zuthal @Li @fizzyizzy05 @0x4d6165 Every time I see folks say "screen readers should just handle them as letters" it dismisses the fact they all are useful characters that mean things in other contexts, but the idea of "in certain circumstances" is the first time I've read a related take that feels vaguely viable. I could see some folks going "aha! A use for LLMs!" but someone on fedi wrote a handcrafted "micro language model" -- his words analogizing it -- so my stream bot could detect when someone key smashes and automatically play the "password generated" event I've got set up for it. Extremely low CPU, could detect quickly in this case. But that all still requires developers of these readers to implement functionality based on this misusage. Where do you draw the line? What about the folks who use those characters from non-Roman written languages that look like Roman characrers, like the racist "Asian font" generators?

Replies (1)

  • @zuthal@floofy.tech 2026-04-05 15:40

    @KayOhtie @Li @fizzyizzy05 @0x4d6165 I think they should, without regard to content or intent, parse anything that a reasonable sighted reader would parse as being intended as funky-looking latin text as latin text. The one that would I think be the hardest one to parse correctly is "faux cyrillic" since there there would be some overlap between that and misspelled real cyrillic, and so ambiguity which to parse it as.

    Open ##1225396