Engineering for Every Modality
Transcript
Made automatically from the video's sound, so it may contain mistakes. Click a line to play from there.
Welcome back. Let's dive straight into this explainer and blueprint for season 10. We're bridging directly from our visual design principles straight into full stack engineering for absolutely everyone.
If you're a product designer, a front-end or mobile engineer, a PM, a QA lead, or yes, even if you're steering AI agents with unguided vibe coding, this is your foundation for building interfaces that actually work in the real world.
Beyond the legal checkbox, the web aim reality check, tackling the five modalities from broken to fixed, valuing lived experience over checklists, and finally the 10-minute smoke test. Let's get into it.
Beyond the legal checkbox, engineering for equivalent independence. So, what are we really talking about when we say digital accessibility? It means empowering users with disabilities to operate interfaces with the exact same dignity and the exact same speed as anyone else.
It's definitely not just about ticking off Section 508 or whack-ag legal boxes on a compliance spreadsheet. Think about it. Imagine trying to just do your job or buy a present for a friend online and being totally blocked by a fundamentally broken interface.
We are engineering for equivalent independence here. To get there, we rely on the USWDS and W3C guidelines as our foundation, specifically the four core principles known as poor, P-O-U-R. Your interface has to be perceivable, operable, understandable, and robust.
Meaning users must be able to actually perceive your information, operate the interface without being forced into impossible interactions, understand what on earth they are doing, and trust that the underlying code is robust enough to work with their specific assistive technologies.
The WebAIM reality check, and exactly why unguided vibe coding fails. The sheer scale of the problem out there right now is just staggering. Take a look at this number. Over 56 million distinct accessibility errors were detected on just the top 1 million home pages this year alone.
To put that into perspective, that is an average of 56 barriers on every single page, literally blocking real human beings from participating in the digital world. And honestly, what's even more alarming is that homepage complexity is just ballooning.
We're now averaging 1,437 elements per page. That's a massive 14.3% jump in a single year. A lot of this is being driven by our reliance on third-party frameworks, unguided AI defaults and vibe coding.
I mean, why wouldn't you trust an AI to instantly build a complex, compliant interface for you, right? Well, because while these outputs might look visually fine on the surface, underneath, they are often fundamentally broken.
This is exactly why we cannot rely solely on automated scanners. As you can see, they only catch a tiny fraction, maybe 30 to 40%, of real barriers, completely missing critical contextual failures and leaving users stranded. And relying on AI or vibe coding defaults.
That yields near 0% compliance, because they generate entirely inaccessible ARIA labels and trap user focus. The absolute only way to secure equivalent independence is through human testing with actual assistive technology.
That's how you hit that 100% barrier identification mark. Alright, part three. Five modalities, going from broken to fixed using our two recurring apps, pathway and operations desk.
Let's start with the visual modality. Failing to use high contrast tokens or explicit text leaves users completely guessing when colour alone tries to convey an error. Look at the broken state on the left.
Pathway's washed out pastel button is basically invisible to low vision users, and operations desk just slaps a red border around a broken field. Now look at the fixed state on the right. We resolve this tension by applying high contrast tokens, explicit error messages, and clear icons.
Meaning should never, ever be communicated by colour alone. Now, for keyboard users, trapping focus or skipping elements is an absolute disaster. This is why engineering native button controls and proper escape focus returns is so crucial.
In our broken example, hitting the tab key completely ski over the lesson cards in pathway because they were lazily built with meaningless div tags. Over in operations desk, the user's focus gets completely trapped behind a modal drawer.
To fix this, you absolutely have to use native semantic controls, ensuring that your tab, shift tab, enter, space, and escape keys actually behave predictably. Imagine listening to the broken verbatim screen reader speech on the left. Clickable, clickable, button, graphic icon underscore 24, link.
How frustrating is that? Now compare it to the dignified semantic clarity of the fixed output on the right. Main landmark, pathway course catalog, heading level one. Start lesson two, landmark regions, button.
We absolutely must write code so that screen readers can quickly orient the user with actual useful meaning. What happens when low vision users zoom in to 400%? Well, responsive design has to kick in with fluid containers and single column reflow to completely eliminate horizontal scrolling.
On the broken side, Pathway's sticky video toolbar clips right over the transcript, and operations desk forces the user to awkwardly scroll left and right just to read a single sentence. It's like a terrible game of ping pong.
On the fixed side, we use WACAG standard REM fluid containers, proving that a truly responsive layout adapts perfectly to immense magnification without forcing that awful typewriter reading experience. Finally, let's look at mobile and voice control.
Leaving tiny 18 pixel icons in your UI leads to mobile misstaps and totally breaks voice commands. If a user on speech recognition says click save, but your icon doesn't have a matching accessible name in the code, they hit a total dead end.
Expanding those touch targets to a minimum of 24 by 24 CSS pixels and ensuring the visible label actually matches the code label changes everything. Lived experience over checklists and the importance of co-designing with disabled contributors.
As the W3C and gov.uk standards so perfectly remind us, a checklist replaces the person. Real inclusion requires evaluating with actual users. You simply cannot automate empathy.
You can't rely purely on automated scanners to tell you how a complex interaction actually feels to someone navigating without sight. This is why paying a blind engineer or a deaf UX researcher for code reviews is an absolute game changer.
It exposes real-world dead ends that no automated test will ever catch. When you watch a blind screen reader engineer demonstrate high-speed landmark navigation on your app, you realize instantly that you are designing for power users with immense technical skill.
Designing with people, rather than just for people evaluates true usability far beyond mere technical compliance. Which brings us to our final chapter, part 5, the 10 minute smoke test, featuring annotated code repair and audits.
Just look at the difference a single semantic swap makes in the code. On the left, clicking a fake div button triggers a silent, vanishing toast notification that a visually impaired user will literally never know happened. But on the right, by simply swapping to a native button and an ARIA live polite status region, the screen reader politely announces saved lesson two progress.
It's such a tiny code change, but it grants the user total independence. So before you commit your code, I want you to lock in this 10 minute, 5 modality smoke test. Unplug your mouse entirely, zoom your browser window to 400%, turn on voiceover or NVDA.
It takes just minutes to feel the friction for yourself. If you can't navigate your own form by unplugging your mouse, or if your layout entirely breaks at 400% zoom, your users don't stand a chance. Which brings us to the ultimate question for this explainer.
Are you ready to abandon unguided vibe coding and actually engineer interfaces that work effortlessly for everyone? Will your current AI defaults survive the 10 minute smoke test? Really think about that.
Next time, we'll begin our deep dive into the real human beings behind the screens. In episode 1 people, barriers, and assistive technology. Thanks for joining me for this blueprint, and I'll see you in the next one.