🇨🇦Samuel Proulx🇨🇦

@fastfinge@interfree.ca · Joined ⁨Jul⁩ ⁨2026⁩

completely blind computer geek, lover of science fiction and fantasy (especially LitRPG). I work in accessibility, but my opinions are my own, not that of my employer. Fandoms: Harry Potter, Discworld, My Little Pony: Friendship is Magic, Buffy, Dead Like Me, Glee, and I'll read fanfic of pretty much anything that crosses over with one of those.
keyoxide: aspe:keyoxide.org:PFAQDLXSBNO7MZRNPUMWWKQ7TQ
xmpp
fastfinge@im.interfree.ca
keyoxide
aspe:keyoxide.org:PFAQDLXSBNO7MZRNPUMWWKQ7TQ

Replying to @⁨wronglang@bayes.club⁩

@wronglang @menelion @matt @nolan @zersiax Even something as simple as reading logs, establishing what normal values are, then collating and describing periods of time where values go outside of normal could help. Just knowing quickly that at the exact time this one endpoint is called disc io goes way up and database response time goes way down but cpu and ram stay within there normal ranges could help a blind human pinpoint where to start looking.

Replying to @⁨wronglang@bayes.club⁩

@wronglang @menelion @matt @nolan @zersiax You’re correct as far as that goes. But audibly representing large amounts of graphical data is still an unresolved research problem. If there was one true proven solution to that part of it, I’d already be pushing hard for the tools to implement it. Whereas AI skips the problem entirely (when it’s not just making things up) by pinpointing and describing the small slices around the problem exactly (“Whenever anyone hits endpoint X, database Y freezes for X time because storage array W is overloaded”).

Replying to @⁨matt@toot.cafe⁩

@matt @nick Ah, got it. I misread you as suggesting yet another accessible programming language. I think the problem with Quorum falls less out of it being accessible, and more out of it being an “educational” language. Educational languages either never escape into the real world to help students do anything useful (scratch, logo) or they do escape into the real world and set everyone back for years (basic).

Replying to @⁨wronglang@bayes.club⁩

@wronglang @menelion @matt @nolan @zersiax Yes. But this smells like a problem that requires AI to solve. Or every program ever to write better logs. As it is, how do I accessibly reconcile my web server logs, my database logs, and my storage activity, so I can track all the symptoms of an issue? Sighted people get graphs with colored lines from stuff like grifana or whatever. Blind people get four log dumps in four different formats, without even colored lines to help us, and told to just listen to every line read out.

Replying to @⁨matt@toot.cafe⁩

@matt @menelion It’s not about the language itself for me. It’s about how inaccessible the tools surrounding it are. Debugging is still awkward and terrible. Designing forms still sucks. Profiling is a pita. We can’t test on the iOS or Android simulator last I checked. We have nothing as clean and easy as unity. All tutorials and cookbooks are incomprehensible YouTube videos, LLM slop, ad covered, or all three. AI allows me to use one interface, with one modality, to do the hard bits.
Isn't interesting that before #AI could code, assistants were always gendered #female (Siri, Alexa, etc.) As soon as AI became able to code, it got #male or non-gendered names (Claude, Chat GPT). I'm not saying explicitly gendering assistants as female was a good thing. I'm saying it's interesting that the trigger for tech/society in general to stop doing that was when the AI became able to write programs. It really does say something unflattering about our assumptions and worldview.