A guide to AI prompting from a blindness perspective

This article was originally written for the Google Developer blog. However, it's been a number of months since I've heard from them, so I'm publishing it here.

stuff.interfree.ca/2026/09/05/ai-prompting-from-a-blindness-perspective.html

The Problems and Successes of Fediverse Comments on my blog

I've been running fediverse as the commenting system on this blog for a while now. Here's how it's been going. #20260822_BlogCommentsViaHashtags

stuff.interfree.ca/2026/08/23/fediverse-comments.html

Also @nprofile1q... I think I might have found a bug in the way you do OMEMO:2. You're doing prekey via Curve25519/Montgomery + XEdDSA. But the specification requires identity keys are encoded per RFC 8032 (Ed25519) while prekeys are RFC 7748 (X25519). I may be confused, however.

If you want to communicate with your Hermes Agent via #XMPP, here's a plugin for that. Code is extracted from three different PR's to the hermes agent github repo to add XMPP that weren't being merged, and moved into a plugin. I just cherry picked and packaged everything with the help of local models running on my laptop, with a final pass by a cloud model (if you read the readme carefully you can tell what cloud model) to update documentation and fix a few loose ends the smaller (but free and more environmentally friendly) local models left. Everything I could test is working. Unfortunately I don't have an XMPP client that supports reactions, typing indicators, threads, or voice messages. But it connects to the server via a secure connection, and both versions of omemo work. However you feel about #AI, communication should at least be E2E encrypted and not controlled by discord/telegram. github.com/fastfinge/hermes-xmpp-plugin

The State of Modern AI Text To Speech Systems for Screen Reader Users: The past year has seen an explosion in new text to speech engines based on neural networks, large language models, and machine learning. But has any of this advancement offered anything to those using screen readers? stuff.interfree.ca/2026/01/05/ai-tts-for-screenreaders.html #ai #tts #llm #accessibility #a11y #screenreaders

That feeling when I write documentation for something, then find the documentation I forgot I wrote for the same thing two years ago, and the wording and structure are almost identical. Right down to the same silly jokes. Then while editing I noticed a mistake I made in this version, and the exact same mistake was in what I wrote two years ago. Am I just an #llm that's made out of meat? Never mind anyone else: will I even realize when I replace myself with an #AI? Maybe it's already happened...

I don't usually share #work stuff here. But this may be of interest to #blind and #visuallyimpaired folks who live in the #UK: rblind.com/post/8599526 #accessibility #screenreader #a11y

I don’t usually share my work #a11y stuff here for various reasons. But I’ll make an acception as this article might be of interest to some of my followers here: makeitfable.com/article/mobile-usability-reality-check-for-inclusive-ux/ #mobile #accessibility

Hey, #fediverse experts: I have a question. On my blog, I embed comments from the #fediverse. I also link to the comment thread on my instance. However, nobody else is using my instance. Is there a service where I can pass it a URL to a note, and it will ask the user what instance they use, then direct there instance to go get that note so they can respond? It currently looks like this, comments at the bottom; I'd like to replace the "this post on the fediverse" link with some way for users to pick their instance: stuff.interfree.ca/2025/05/02/sleep-solutions.html

Today in #ChatGPT attempting to be helpful with a thing I'm building: "Consider bumping the font size in your CSS so NVDA/VoiceOver don’t make you squint your ears." I, uh, I'm not sure #AI is ready to do #accessibility yet. I mean, I should bump the font size for #a11y reasons...those just aren't the right reasons. I can't wait until we have completely #blind designers building CSS with nothing but AI. Anyway, back to squinting my ears!