Accessibility panel

Future Frontend 2026 (Accessibility Panel) / 12:41

Watch this talk on YouTube

Transcript

13 paragraphs

This is an automatic transcript of the recording above. It is published in full and unedited, apart from correcting names the recogniser reliably mishears. It will contain mistakes.

00:05Um hey Daniel. Uh first question is for you. Uh This is very >> [laughter] >> a very serious question. Um what's your favorite chocolate brand? >> Uh cool. Let's uh shift gears. Oh, there we go. Cool. Uh so, this is actually I break this into two parts. Uh commercially available chocolate, I go with the classic Feodora blue. Uh you can get it anywhere. It is my favorite chocolate that I can just have on a daily basis. But then there is a chocolatier in a country called Sao Tome and Principe. It's a small little two couple islands on the equator off the west coast of Africa and they make the greatest specialty chocolate in the world. And I would be happy to uh have more of that. So, yeah, Feodora and then Sao Tome and Principe.

00:59>> Cool. Good answer. Um let me see. Um the next voted one uh let's ask that one. Not sure who that is for, but uh maybe one of you. I've noticed AI made UI many times taste the same. I don't know if it is the same color world they use or something else. Once more people realize this, do you think creators will try to hide they used AI for the UI? >> I'll take a shot at this, Tejas. >> Yeah, I think I I'll change think I hope they hide that they used AI for you because I I don't like looking at it. Like I I I was saying earlier, I spoke at a conference some time ago. It was an AI conference and I was MCing. I was doing your job. And and every speaker had the same like aesthetic to their slides. It was all like a light orange background with orange headings and this like serif font. And it looked like Anthropic's brand, right? And and so I was like, that's And as an MC I called it out. I was like, is it just me or did the past five talks all have the same aesthetic?

02:09And so I think a lot of people are just creating things with AI. The problem is they're not applying their own personal taste to it. And I think if they can if if as part of trying to hide that they used AI, they just actually think for themselves and apply their own taste to it. I'm I'm all for it. Let's do it. In fact, if you look at the movie industry, right? Like it got to a point where there's really we got bored of the movie formula. Like a lot of the Marvel movies they follow the same formula and we're kind of like bored of it. And so what the movies that are doing well are these really novel unhinged weird obsession backrooms Everything Everywhere All at Once Parasite. Like these movies have like weird energy to them. And by weird I mean they're novel. And as human beings we're we're born and raised novelty seeking. And so I think novelty is a beautiful thing and and I think if if we end up hiding that we used AI assistance, but then we hide it by covering it up with our own novelty and our own beautiful creative genius, then let's do it.

03:10>> Yeah. Yeah, great. And feel free feel free to chime in at any point. A lot of these are for Tejas. Trying to look at trying to get everybody in here, but oh another question. Maybe maybe maybe we'll do the favorite bands of all of you. The question was Daniel, what's your favorite metal band? But but I'm going to ask all of you something similar. >> Uh as someone living in Finland I'm embarrassed to say this, but my favorite metal band they're called Mister Misery from Stockholm and I'm sorry, but they're my absolute favorite. That's fair you did. >> Cool, cool. What about you Ramona? Favorite any band? You a metal fan? >> Yeah, also. Either metal or techno basically. And there's one band which combines both and it's local to me. So, one of my like local bands, Electric Callboy. If you guys know them.

04:02>> Oh, yeah. Yeah, I heard of them. Yeah, cool. They just They're a band >> Yeah, uh it's not metal. It's more like rock. It's a band called Switchfoot. Yeah. >> Cool. Love Switchfoot. >> I've been hearing that name >> forever. >> Yeah, I'm old. >> Is this the Christian band, right? >> Well, they just do regular music. >> Oh, okay. Now I'm getting confused. Okay, yeah. >> No, they're from California, though. Yeah, yeah. >> Okay, okay. >> Yeah. >> Okay. Oh, this this has so many votes here. Um >> It was a bot man, for sure. Somebody's like spamming. >> I know, right? >> [laughter] >> Um so, how much do you trust the results of or conclusions your agents come up with? >> I I don't I don't trust it. But, I do trust myself.

04:45And so, I build a harness. You may hear this word harness. I I did a talk about this recently at an AI conference about what does harness mean? How do you build one? I built a harness around my agent which I trust. I trust that with my life cuz I built it myself. And the harness an agent harness is exactly like a harness you would give to like any animal like a dog or something. You know, a harness is like a leash. So, I trust my harness. I don't trust my agent. >> Yeah, cool. Good answer. Another one for Ramona. Um Press that. So, what is considered an accessible alternative to CAPTCHAs? >> So, my favorite one would be the Google one, the reCAPTCHA, which is basically a CAPTCHA which cannot be like or will not be that much noticed by the user, but it's like the reCAPTCHA judging the behavior of the user inside of the web. That could be IP, that could be like click patterns and I'm not sure if they completely disclose what behavior they're judging, so that an attacker or something cannot just imitate it. But, it's a risk-based approach, so the user will not notice, but they will be judged, and if the user does notice it, it's not like a memory test. So, I really like those invisible risk-based approaches. Doesn't need to be by Google, but I guess that's the most common well-known one.

06:09>> Cool. Marcus done, and let me just kind of figure out this app here. Okay, um maybe a question for all of you. Uh this is really good. Uh like so, how do you how do we approach safety in the era of AI? Want to take a shot at this at this, Daniel? >> Safety. Uh well, at least I would consider like accessibility to be, you know, safe safe usage of the web. So, I think um learning more about agent harnesses and stuff, uh as agents have mentioned, will help get us there. Um I know Google is working on or releasing uh different skills to keep agents up-to-date on latest features and evolving as the web evolves with it. So, um finding a way to essentially not let the context of the agents go stale, but also not trusting what the agent gives you and having a harness to make sure that it stays, you know, not creating flashing UIs, creating, you know, accessible login forms, and and all that stuff. So, um manual review, harnesses, and uh current context.

07:22>> Yeah. >> When it comes to safety as of security, I would um recommend you to take a look at the OWASP stuff as well, because they do have a top 10 for LLMs, which is super interesting and gives lots of advice. And generally, when it comes to mindset, I like to phrase it as being professional pessimistic, if you can call like that, so basically having an intuition and really being careful when it comes to working with AI-generated code or having your let's say chatbot reacting to prompt. Be aware of prompt injection. Just try to educate yourself and act accordingly when it comes to all AI usage. >> Nice. Cool. Um any takes on safety to >> Yeah. Um I think it I said it before with the harness is is just really invest deeply in a harness. I think that's the that's the move right now.

08:19>> Absolutely. Um and this came actually and this is this uh directed pages. Um but I remember somebody asking this uh last year and it actually it was after um Rachel's um talk because she kind of touched on on kind of like a a was it like web without browsers or um but um if agents use the web more than people do, the value of ads will inevitably fail or fall, sorry. Uh which will make the web less profitable and more harsh work environment. Thoughts? >> Yeah, good I like that this is just an open-ended question. Thoughts? I I have I have many thoughts. I think if we think the web exists for ads, we have a very short-sided view of the web. The web Let me remind us where the web came from because the I I'm not sure the person who asked this is aware. The web was invented at CERN uh in Switzerland. Uh it was a research project to help connect the internet um which was read and write at the time. It just became a read-only directory of web pages for research. Um and the cool thing about CERN is it was owned by no one. Uh it was not owned by an ad agency and the purpose was not sharing of ads.

09:36It was sharing of information and ideas. Um Tim Berners-Lee, that's that's where he invented it. That's the vision and that's what it was for. Ads ruined the web. In fact, I will say this, um ads uh makes the web maybe profitable, but it also makes the I think ads make the web a more harsh work environment. And so, the absence of ads, uh I think is great. It's maybe not as sustainable without ads. Uh and and we'll have to explore like newer models for revenue. Um but I I I don't think that's going to be a problem because we've always been good at that. I think even now, for example, um you have to pay to use something like Perplexity uh or ChatGPT. ChatGPT has a free tier that's kind of useless. Um and so, when you pay, you're already like nobody pays for google.com, right? Except you pay with your data for their ads. Um but if if it could happen that you just have to pay some type of monthly subscription or something um for a level of usage beyond some base tier.

10:33If that means less ads and a more pleasant work environment, I I can do that. And I think there's companies that would be willing to fund it. So, there's no definitive answer for this. I think it's too early, but if ads and their value falls, then we'll find a way. Um but the web was never about ads in the beginning. And I don't think ads are kind of the end as well. >> Yeah. Good answer. We'll do one final question for Ramona, um cuz it's getting late. Um how do we explain to management um that we are spending time on tuning accessibility if they don't see the difference? >> That's an interesting question, actually. So, first thought is maybe making it visible. There's some tools like the accessibility testing one, maybe even like ours.

11:19And then trying to use this as a like argumentation line to say like you can see what effects it has not only when it comes to people who are dependent on the but also when it comes to search engine optimization and stuff like that, because this place in as well, and then you have a price tag to it. And apart from that, you could use like a screen reader and stuff to make it visible. And if you don't want to go this like proving it rude, I would say like, "Okay, try to remind them on points in time and their needed accessibility and tell them, "Yeah, okay, would you be the one to basically exclude all of those people?" I guess that's a pretty strong argument to go for accessibility, because you will exclude people if you don't go that road.

12:18And as a last resort, the laws. >> Absolutely. Great answer. And I think that's a wrap for today. Thank you so very much for your talks. Really informative and um thank you everybody for sticking to the end, and um this was a this was day one. Um let's give a round of applause to our speakers. And to the MC. Yes.

More talks

All 23 transcripts

Elsewhere

There is every talk I have given, all 80 of them, ConTejas Code, the podcast, and Fluent React, the O'Reilly book on how React works inside.