From Digital Curb Cuts to Alien Intelligence: Best of 2025

Transcript

DAVID CARTY: As AI reshapes our digital experiences, the function of quality assurance has never been more complicated. Over the past year, we’ve had the great fortune of interviewing guests across many areas of domain expertise. Their insights are helping to form the digital quality strategies of tomorrow. For brands trying to stay ahead of the curve. From AI to AR, designing to developing, quality is a full life cycle, full team priority. And on the Ready, Test, Go. podcast, we will strive to provide the kind of insights that you can’t get anywhere else. Without any further ado, let’s jump right in.

This is the Ready, Test, Go. podcast brought to you by Applause. I’m David Carty. Before we move forward in 2026, we wanted to take a look back at some of the key insights from our guests in 2025. And if there’s one thing that should be on your list of goals for the new year, it’s prioritizing the end user. In the rush to launch new features, it’s easy to forget that for many people, the digital world is still full of barriers.

Digital accessibility is becoming a global imperative, and two of our guests offered sympathetic perspectives on improving usability, not only for people with disabilities, but for all users of their products.

JAMILA EVILSIZOR: 90% of home pages are inaccessible. That term “inaccessible” means they don’t work. They don’t work for people. I was asking a user naively, how do we make this experience more delightful? And their response is, I am just looking for it to work. I asked, is there a competitor out there that really makes this experience shine? Once again, the answer was nobody really has this figured out yet. No one has it working. It almost reminds me of Maslow’s hierarchy of needs. You have to get the basics down before we can start talking about things like delight. That’s where we are right now, unfortunately, in the digital landscape, is just trying to make sure that we have a system that works for everyone so that they can just accomplish the one simple task they were doing, which just means that we need to be doing better across the board.

Accessibility is about making it usable for a data analyst with a migraine, a court reporter with carpal tunnel, an engineer with dyslexia, an executive with color blindness, a customer service rep with a speech disorder, a deaf graphic designer, for you as you age. We’re building things for our future. Another misconception is that a person with a disability always has an assistant or a family member to help them out if things get rough. Some might, but not always. And even if they did, I want you to think about what you’re asking of a person. If our app doesn’t quite work out, they can always ask for help. Think about your day to day. Do you want to have someone sitting behind you helping you every single moment of the day click a button, read what this graph says, make sure that your focus isn’t trapped with your mouse? Everyone wants to navigate this world independently, with autonomy, with dignity, with privacy. And a medical diagnosis does not make that any less true.

WILLIAM RUESCHEL: There’s actually a concept called a digital curb cut in the US. They’re called other things globally. But if you’ve got a sidewalk and you’ve then got a parking lot, and the curb cut is a ramp that allows you to transfer between the parking lot, the lower elevation to the higher elevation on top of the sidewalk. So crosswalks, they all have curb cuts. That is because we need to allow access for people who are using wheelchairs. That’s why those curb cuts exist. But the benefit goes beyond that.

Think about people that are riding bikes, people who are pushing strollers, people who have shopping carts. All benefit from having those curb cuts. So it’s an ADA requirement to have them, but everybody benefits from it. So that innovation really improves everything. Digital curb cut are features that get built with accessibility in mind to solve a problem for maybe somebody that has a disability and is using a screen reader, but that benefits everyone else. So some of those famous examples are things like captions. Captions are for people who are deaf or hard of hearing that can’t hear speech in synchronous media, but they can read it. Captions are used primarily by people who can hear the speech, but who are in a situation that is not an easy thing to do. For example, you’re in a sports bar, and it’s noisy and you’ve got 14 TVs. You might have captions on so you can still see what the announcers are saying. Or this is also a tremendous benefit for language learners. 80% or more of captions users are not for people who are deaf or hard of hearing. They’re because people benefit from having those captions or subtitles.

You already mentioned contrast. Having brighter screens, higher contrast allows you to use your device or your phone or your app in more locations. If you expect your users to have an app on a phone and not just be sitting at their desk in office lighting, holding their phone at arm’s length, in ideal scenarios, they’re going to be out in the world doing things. So they’re going to step out into the sun. They’re going to be in shadow. They’re going to have glare. All these things are going to be happening. And so having a readable display, a readable, app readable UI benefits everyone.

CARTY: Have you ever attended a test party, or have you ever experienced the lizard man’s constant? Well, if you have a traditional staged assembly line view of software testing, you might not be embracing the creativity needed to succeed with a modern digital quality approach. Our next slate of guests leans into some of the non-traditional methods of validating high quality digital experiences across the entire organization.

MICHAEL BOLTON: My favorite bugbear is the focus on artifact-based testing, focused on the idea of the test case. No other investigative, cognitive, socially focused intellectual craft uses cases like the testing business does. Journalists don’t use journalism cases. Researchers don’t use research cases. Parents don’t use parenting cases. And for heaven’s sakes, managers don’t use management cases. And developers don’t use development cases either. There’s a reason for people to think of testing in this way, and that is, it is very convenient and very indisputable in a way, very eligible to use test cases to have a nice, tidy procedure that says do this. And then observe this specified, presumably desirable, result.

If we don’t get the desirable result, and we’ve done the procedure just right, there’s likely to be a problem in the product. But there’s an asymmetry in there, and that is, just because we’ve done that, that doesn’t mean there’s no problem in the product. And what we worry about is over-focusing testers on the procedure and on anticipated, as I say, specified, presumably desirable output. Nobody else works that way. Teachers don’t work that way. Even though teachers have multiple choice tests, when we’re getting serious about teaching a kid something, we evaluate the kids learning by– not by giving them something routine, but by giving them a real challenge. When we are evaluating university students for Master’s degrees and PhD theses, and so on, and so on, we don’t give them a multiple choice test. And testing with test cases often amounts to that.

Now, there’s some wiggle room there. There’s lots and lots of notions of what a test case is. Sometimes people refer to it as a test case has a kind of looser structure than that. We would call that a test idea framed around a bunch of test conditions, maybe, because we do want to be able to reproduce problems if we encounter them. We want to be able to analyze them. So relaxing the focus on test cases can make debugging more difficult sometimes, because there’s factors that we don’t track, that we don’t necessarily understand, that we’re not necessarily aware of, even. That’s true for test cases, too, a lot of the time when we’re trying to figure out what’s going on.

GOJKO ADZIC: Got Alexander’s blog post on lizard man, and he wrote about this kind of demographic research study they’ve done where they’ve taken some demographic data, and then they try to do some psychological research. And they were incredibly surprised by people just giving ridiculous demographic answers. There was a number of people that were selecting American as gender, people who were selecting Marshall as nationality or all kind of weird things like that. And then he goes and compares them to a bunch of other things and put this 4% constant that says, basically, 4% of the people in any large population are just going to do things that are not reasonable to you. Reasonable to them, but they’re not reasonable to you.

And so this number 4% keeps coming up in different places. And he just chose to call it the lizard man’s constant. And when I read that, it was mind blowing for me, because I realized I’m approaching this thing from a perspective of these people are crazy, or these people have no idea what they’re doing, or they’re incapable, or they’re malicious, but they’re not. Of course, there will be some percentage of malicious people trying to abuse the system, but mostly, we’re struggling because there’s a mismatch between their intents, their capabilities and what the system provides. It’s up to us to understand the lizard logic. By figuring out what they are actually trying to do and what they want to achieve, we can then figure out, is this a good way of supporting, and do we have a good way of supporting it? Should we even be supporting that? You learn to tell people, I was using a system that’s going to be having this operational awareness of what people are doing to your system that’s maybe not what you intended.

LISA CRISPIN: If the developers don’t have a lot of testing skills, or they really think somebody else should be doing it, have ensemble test sessions, or you can rebrand them as test parties and say, hey, you’ve got a new feature that you think is going to be ready to deploy soon. Let’s spend 30 minutes doing some testing on that. It might be exploratory testing where we have charters. It might be ad hoc testing. Invite the developers. Hey, you want to get together for 30 minutes, and let’s just try out this feature and see how confident we are about it. And in that session, of course, questions come up. Most of the time some issues are found and again, they’re– phew! I’m glad we found that we were about to put that in beta. We better not do that.

And so they see the value and they start to understand what is this weird testing thing all about. And doing it earlier is better. And maybe we should have some of these conversations before we even write the code so that we have a shared understanding of what to build. And the most frustrating thing is to deliver a story to your product owner because you think your team did a great job on it, and the product owner says, that’s not what I wanted. So somehow, you estimated that story and worked on it without actually understanding what it was, and now you got to do it again. Nobody wants that.

And so anything you can do to save time by having conversations and using structures that really help us, like identifying risks, something like risk storming that you can do in an hour, and look at what are our most important quality attributes. What are the risks that might happen that affect those quality attributes in a bad way? How are we going to mitigate those risks? It might be with testing. It might be with making sure that we log the right events and data so we can monitor it, have observability. So having those conversations early, it’s a little investment of time. Example mapping at the story level where each story, what’s the goal of the story? What are the business rules? For each business rule, give us some concrete examples of how it should behave. The product owner or stakeholder can give us those concrete examples. We can turn those examples right into executable tests. So the code is written. Once the tests pass, oh, we’ve got regression suite now. It all hangs together. And again, I talk like it’s so easy. It takes a lot of time to take a team that doesn’t have those capabilities and start building them in.

CARTY: Everyone is aware of the seismic shift that AI has caused in the digital marketplace. The consensus from our guests is it is moving fast, and the playbooks have already changed.

Some claim that in this environment, the biggest risk is not moving too fast, but moving too slowly. Yet others argue that deeply human ideas like ethics, responsibility and experimentation must factor into AI-based products to ultimately benefit society. Here’s why brands need to innovate not only in their design and engineering approaches, but also in their digital quality strategies.

ANDY SACK: I probably would say that everybody’s moving too slowly, and right now there is no risk of moving too fast. I don’t think you can under-invest in AI right now. I say that because AI is moving so much faster than all of us. Maybe there’s some risk that by moving fast, you might get locked into a model. But the fact is you should be building, knowing that the models are going to be better tomorrow than they are today. There’s not going to be a worse model than there is today. And what I see in the market is, basically, people know AI is a big deal, but across the board, 70% still think of it like a Google search replacement and don’t really understand that this is not software. This is some form of alien intelligence that has been created. It’s an amazing set of capabilities. And so I think the risk is actually not moving fast enough. That’s where I come down. Now, there’s lots of implications from that around quality, around risks, and opening yourself up to fraud, et cetera. So I don’t want to minimize those risks, but I think the risk is much more on the other side.

NATHAN CHAPPELL: Responsible AI is not the same thing as beneficial AI. And so just saying something’s ethical doesn’t mean that it actually is avoiding long-term consequences. Social media, for example, is not unethical as a technology. In fact, it was actually designed to essentially connect people that wouldn’t know each other otherwise, and really, ideally, bring us closer together. But of course, the net effect in how it’s been abused, essentially, has been quite the opposite. So increases in anxiety, and depression, and teen suicide, and all those net effects of too much. And so, while social media is not unethical, it’s also not beneficial to the long-term of humanity.

I think that’s something that organizations, both for-profit and nonprofit, every organization that is selling a product, or a service, or whatever, needs to really wrestle with, again, ensure that your AI is not unintentionally creating bias. We don’t even talk about bias anymore. That was such a big topic several years ago, but we got really good in predictive AI at understanding how models are making decisions. So now, what used to not be transparent is now totally transparent. I’ve become a huge believer in transparent AI, which means that AI can be interrogated. So predictive AI allows you to interrogate it, which means if I’m going to make a prediction that you’re going to buy a cell phone, you would get a score in the likelihood of you buying that cell phone. I should be able to tell you, what math was used for you specifically that led you to a score of 80 that you’re likely to buy a cell phone? And predictive AI now, compared to even seven years ago, allows you to do that completely. There is no reason why you can’t actually interrogate predictive AI. Generative AI, on the other hand, very different. Generative is this crazy black labyrinth. In fact, OpenAI, and Anthropic and Perplexity, there’s not a lot of clarity, even from those organizations, why those models work as well as they do. Now, we have to trust that they’ve taken lots of steps to mitigate bias in those models. But we know that there’s bias.

SANDY PENTLAND: Let me just point out the history of certain sorts of innovation. So let’s just take physics, mechanics, how you build things out of steel and so forth. So Newton had his theory, but they had to completely rewrite that theory three separate times before they were able to build bridges that didn’t fall down, because the way that he had posed it didn’t work in certain situations and had certain hidden boundaries. So even something like, how do you build a bridge, turned out that we couldn’t do it a priori. We couldn’t reason it out. We had to actually build some bridges, have them fall down and fix them. So now physics is an incredibly well defined area where we understand the bejesus out of it.

So if we can’t even get things right there– and notice that we still build models, and crash them and try it. Even though we have theoretically all the physics, how are you going to deal with the real world where there are memes, and different cultures, and the economy changes, and other technologies? No, things change. And we talk about this as these four horsemen, four major errors people make in terms of not noticing that what they’re doing used to be right, but now is wrong. That’s the core problem, is that you don’t know everything. You don’t know all the trends. You don’t know where they’re going. We’re not very good at predicting them. If we were, hedge funds would be a lot better than they are today. And so you’re in this sort of churning mess of an ocean of technology and society. And you have to not think that you know which way is forward. You have to continually be on the lookout, testing things, not expensively, but cheaply so that you know which way to turn when there’s a leak in the boat and things like that.

TOM EMRICH: Yeah, AI is becoming like a peanut butter to the immersive jelly, as it were. And first off, AI is really the underlying technology of augmented reality. Augmented reality uses computer vision in order to help make sense of the space, and your face, and your feet in order to make the magic happen. So they’re definitely not at odds, although it sometimes in the media, it makes it seem like we have to choose between one emerging technology to the other.

Specifically, I think with generative AI, I think there’s a couple of really interesting opportunities that are bubbling up. The first is in helping to create assets. And so we’re seeing this, of course, with text, and video, and GIFs, and photos. But there have been great advancements in the use of generative AI to create 3D models, 3D animations, 3D meshes, and also to use the video and the photos that are being generated as textures to apply on these 3D meshes. And so it’s still somewhat early days for 3D media creation, but it really is in a good state to get your feet wet with prototyping, in particular, maybe some background trees, for example. And I think it’s just a matter of time. As we look to the long-term future, I think we can definitely see that the writing’s on the wall that generative AI will play a big role in 3D asset creation. And this is really important because at the end of the day, when we’re talking about augmented reality and virtual reality, we’re really talking about the shift of content from 2D to 3D. And you’ll need to have 3D content, models, animations, spatial logic, spatial audio in order to create a compelling, immersive experience. And so having this generative AI assist in this way can really make this experience creation more efficient, and perhaps more within your budget.

The second way is as a co-pilot, allowing for generative AI to be paired with developers to help with debugging. And to help with the coding experience itself could help make this development process much more efficient, a lot faster, and allow it to really ramp up your team’s. And this is another key thing with spatial computing that’s going to require a brand new skill set within your teams, and generative AI can really help with that.

CARTY: To close out our best-of episode, we turn to the new challenges presented by this increasingly data-driven world. With all of this technology at our fingertips, it’s not just about what we can build, but what we should build and how we are operationalizing human data in doing so. Our guests offer a sympathetic and refreshing look at how brands can be mindful of the purpose behind their products. The good intentions of today might just be the regrets of tomorrow.

SANDRA MATZ: So it’s almost like a thought experiment that I really loved, that I heard someone at Apple talk about once. And it plays into this idea of, how comfortable do you feel collecting data? Typically, you’re designing a product and you have a team working together. Everybody’s excited about the product. And most of the time, people really do have users’ best interest at heart. So I think most of the time, when you speak to people, they say, here’s how we’re going to make the product better. For that, we just need the data.

And what they do is they call it the evil Steve test, so back when Steve Jobs was still the CEO. They go through this thought experiment of what would happen if tomorrow we get a new CEO that has completely different values to the value system that we have right now, is not trying to help users, but is just trying to exploit them to the furthest degree possible, as much as possible. And would we still feel comfortable collecting the data that we’re collecting today, and setting up the system in a way that we’re setting it up right now? And if the answer is no, you should go back to the drawing board and see if you can do better. So I think it’s a nice way of putting yourselves into the shoes of your future self, and without necessarily the pure enthusiasm that project teams typically have for their product. So it’s like playing devil’s advocate in a slightly structured and fun way.