Home | Notifications | New Note | Local | Federated | Search | Logout

Note Detail

Reply to @hboon@mastodon.social
Chee Aun 🤔@cheeaun@mastodon.social (2026-10-23 17:25:24)
@hboon technically coding agents can be steered towards improving such issues too, where accessibility affects humans.

From my observation, the surprising thing is that, code spit out by coding agents are technically *more* accessibility-focused - not sure if it's trained to do so. It kinda make sense because it becomes a win-win situation; when a UI becomes more accessible to humans, it’s also more accessible to agents.
---Reply---
Hwee-Boon Yar@hboon@mastodon.social (2026-10-23 17:35:59)
@cheeaun You could steer, or maybe the model becomes better by default, towards writing more accessibility-focused code; but I was wondering if we lose the drive to improve frameworks, standards etc to improve accessibility. Like even if it's harder, as long as it can still be done reasonably with AI, no one might be noticing or feeling that difficulty/gap to improve it.

---Replies---

Chee Aun 🤔@cheeaun@mastodon.social (2026-10-23 17:53:10)
@hboon i think the drive always come from the problem set, then humans fix it either by hand or agents, then it'll be fed as training data to the model, comes back full circle until *all* problems are solved 🤣, assuming that they can all be solved in the end 🙈 (or never ends?)