النص 🔗
speaker-0: People cannot be trusted sometime. I'd like to say helpfully skeptical about whatever ⁓ an LLM spits out. If we run these LLM generated tests ⁓ to verify the state the behavior of the LLM generated product code, do we actually have the information that we're looking for? That's a hard problem. It is a very hard problem. We we all know that code coverage ⁓ At least not when we talk about the quality of the work that we do as testers, as automation engineers, as developers. I think I don't know exactly where and when, but we all have these visions about these big companies that are incredibly successful in what they do and they make a lot of money doing what they do.
speaker-1: Welcome back to Browser Stack Talks, conversations that turn testing pain points into playbooks. I'm your host, David, and today we're moving to tool-driven hype cycles and look at the true essence of test automation. Our guest has spent nearly two decades helping engineering teams around the globe rethink how they get their product feedback. He's an independent consultant, world-renowned trainer, and creator of libraries, but... I can't do this any justice. Bass, please introduce yourself to our guest, our ⁓ followers.
speaker-0: Thank you so much for having me, David. so yeah, ⁓ my name is Bastagstra. ⁓ I'm not going to ask you to try and pronounce that. ⁓ although ⁓ I've heard I've heard people make all kinds of lovely abominations of my last name. I live in the Netherlands. As you said, I'm an independent test automation trainer and consultant. I've been doing this for 20 years, ⁓ to the point where I basically say, well, this is really all I can do. but I've been enjoying it a lot working with individuals and teams and companies ⁓ across the Netherlands and increasingly across the world as well.
speaker-1: Awesome. Well, it is fantastic to have you here. ⁓ Your philosophy is something that's resonated with me really well, is that like you you tend to strip array buzzwords. You tend to focus on kind of allowing people to like not focus on the hype, but go, ⁓ this is really cool. I can do this. I can learn, which is one of the things that I've enjoyed watching you do over the kind of a very long time. and with that, I'd like to kind of dive deep into kind of how you do that. It's like your core principles that you, ⁓ train with people. like, ⁓ you often write that tool, ⁓ that we use tools because we want user you that we always want valuable feedback fast yet many teams fail, ⁓ to do that. They fall into metrics traps. ⁓ they kind of focus in on. vanity metrics when it comes to things like that. So for me, a vanity metrics is like, I've got 100 % code coverage. It's like, great. That doesn't tell me much. How do you actively shift an organization's mindset from kind of test counts coverage to feedback value?
speaker-0: So, first of all, ⁓ that's a hard problem. It is a very hard problem. And and we all know that we we all know that code coverage probably isn't
Load full transcript
the best metric. ⁓ at least not when we talk about the quality of the work that we do as testers, as automation engineers, as developers. but it's so darn easy to count. and to measure and to report on. so one of the things that I like to do is ⁓ show people the the drawbacks or or the shortcomings of things like code coverage. And you could you can just give a very simple example of a test that has a hundred percent code coverage and that fails to detect any kind of actual problem in in the code. ⁓ So that's one thing that I love to do. And as you said, for me, ⁓ the goal of testing in general and and of automation, which is my little corner of the testing world, what I've been working on, is ⁓ providing valuable feedback fast. ⁓ a lot of teams are very good at fast ⁓ at delivering lots of feedback very fast. ⁓ we we write tests fast, we run them. fast. and what I'm so one thing that I've learned to do that I had to learn over the that I had to learn to do over the years is ask these teams well What do you know now? Now that you've written and run those tests and saw all those green check marks or saw those red crosses? What does that tell you about the state of your product? Do you need to know what you need to know now? is this the information that this test suite is giving you? How much do you trust it? are you Comfortable making the decisions that you need to make. And those decisions could be anywhere from let's merge this into main, ⁓ all the way up to ⁓ let's put this in the hand ⁓ in the hands of our end users. so ⁓ If we are going to spend all the time and effort writing, running, maintaining and loving those tests. We need to actually ⁓ make sure that that first of all that that those tests are giving us the value the the the val the feedback that we need. Mm-hmm. And that we can actually trust that feedback. And so can you make the decisions you need to make based on what you know about the product after you run the tests? And ⁓ especially now that we have so much of our code written for us. By these LLMs, the matter of trust comes into play. And that's always been important, of course, because ⁓ people cannot be trusted sometimes. But definitely I I am I'd like to say helpfully skeptical about whatever ⁓ an LLM spits out ⁓ it gives me. so and I think it it's very wise to say, well ⁓ if we run the if we run these LLM generated tests ⁓ to verify the state the behavior of the LLM generated product code, do we actually have the information that we're looking for? And do we trust that information? Do we trust that information that those tests have given us? I think that that's a very, very important question these days.
speaker-1: One of the things you said there that kind of really resonated with me and it's something I've written about and I've spoken about in the past. And I think it's especially important with the age of LLMs and we'll get to a whole section on LLMs later. yeah. Is the, is the amount of overwrite, I think is the way to describe, like that's where my brain went. Is that where people go, ⁓ I'll just throw an LLM at it and it generates. a thousand tests, right? ⁓ And you go, great. ⁓ And like you say, you need to get some information or feedback from it. ⁓ But for me, knowing a thousand tests could take anywhere from a minute to kind of a day to run, depending on the style of test that's created. ⁓ How do you tell, how do you give people the like, confidence to go actually this is the information I need versus this looks very cool right I've burned tokens my manager told me to burn tokens like how do you how do you balance giving the feedback part there like that that to me really resonates it
speaker-0: ⁓ so first of all, I think that a very important thing to keep in mind here is that we as testers that that our business, our role, our responsibility is to provide value. It's not to make the decisions. But it's our role, it's our responsibility to provide the information that others ⁓ could use to make an a a a better informed decision or a better decision in the first place. ⁓ So ⁓ and now I lost track of your initial question.
speaker-1: So I'll re ask it is that like ⁓ in the age of LLMs, you've got a lot of people pushing for a number of tests, they generate 1000 tests, their managers pushing for tokens usage, because that's one of these weird metrics that I've seen being discussed. How do they get? How do they pick a test to get the information that they need versus just having a thousand tests because those thousand tests could take a minute or could take a day depending on the type of test that has been written. How do they, what is the information that they need to surface to make that more valuable? Even if it means I only need one test. Like how do they, they work that out?
speaker-0: so it to me that it is a matter of continually asking that question. ⁓ do you can you make the decision that you need to make, whichever again, whichever kind of decision that is, with what you know about the product right now? If not, ⁓ you'll either need more tests, more testing, ⁓ But you'll probably also need better tests and better testing. And I've done some experiments with this recently. Well, we have these LLMs that write tests for us. And they write a lot of tests very fast. And they sort of kind of look good. ⁓ and once you start actually testing the tests, ⁓ just to see well if they can actually fail and if they uniquely contribute to some form of coverage that is not just line coverage. You'll see that often a lot of the tests start breaking down, which means that they they look good and they look good in the reports and in the pipelines and whatever, but what they actually contribute is very, very little.
speaker-1: Yeah. Yeah. I think that's right. I think kind of the other way. Alan Page mentioned this to me when I discussed it with him, was that ⁓ if a test is not offering user value, user value is the most important thing that a tester needs to
surface. If a test is not doing that, ⁓ it should be removed. And obviously, like, it's quite easy to go, ⁓ I'll remove everything and I'll just have the end-to-end tests. No, that's not what he's saying. It's about like, you know, user values got different layers to it. Like, it's like an onion to check.
speaker-0: Sorry.
speaker-1: Yeah. So one of the things that also that I think you've talked about quite well is that ⁓ You've stated that you don't think a tester exists in a rigid role, right? An archetype of sorts, but rather they're an advocate for quality. ⁓ What does that look like in practice, especially on a highly automated development team?
speaker-0: It's a good question. And in all honesty, I don't think I've completely figured this out yet. but to me, a tester is someone who provides information about the state of the product and and any kind of potential problems that the product might have that would risk the user. or the the the the the pro the development process itself or maybe both ⁓ so there's a whole I know there there's lots of different corners in the software development and the software in the software testing world and you mentioned Alan Page and I've been following following his work for well a very long time as well So he's sort of kind of in a particular corner, and then there's other people who are in different corners. I don't think I've I don't think I've found my corner here yet. And in all honesty, I I think that there's other people who can say way better things about it. To me, again, the the ⁓ I don't really I so I sometimes joke when I talk about this with clients that I don't really care if it's the janitors doing the testing. And yes, that definitely is a good will hunting reference. But but I care about good testing and about again about and and specifically again because I work in automation about ⁓ tests that provide again provide valuable feedback fast. ⁓ and you'll need a little bit of everything.
speaker-1: Yeah.
speaker-0: ⁓ that that requires ⁓ a wide range of skills that requires ⁓ development skills and it requires software testing skills. what I definitely don't recommend ⁓ at teams and organizations to do is is try and find that one person who's gonna do that all and who can do that all really well.
speaker-1: Yeah. Yeah.
speaker-0: So and and ⁓ so that that elusive test automation engineer, software development engineering test or whatever you want to call it. they do exist, but they're very, very rare. The people who can do all of that really, really well. And ⁓ well once you you it's a well the Well, not gonna try and find that one person. It's gonna do it really, really well. It's gonna be heavily dependent on a team and on the context and on their their ⁓ maturity and their their skill level to say, well, ⁓ maybe we have ⁓ a lot of very experienced developers in the team who also are great testers because those definitely do exist. Yeah. And ⁓ that's absolutely fine. And those there ⁓ having a dedicated tester might not have as much value as it would have in a team where ⁓ the developers struggle a little bit with testing for either a a skill or a time or a priority or whatever kind of reason. So a lot of it really depends on the context there.
speaker-1: ⁓ this is awesome. I like it. I'm really enjoying this conversation because it's like every time you say something, it reminds me of a different conversation I've had with someone else who's like, you know, like they've got the first half of your like answer and their answer wildly differs. later, great example of this was Marit. I had her on this podcast. ⁓ and Marit's really great with finding the, Perfect phrase, right? I've never found a person who's so good at finding that way. Like you were saying, you've got these people who kind of have all these skills together. ⁓ and they're very rare, but when they do, they do an amazing work. ⁓ and she, she said like, when you've got that awesome run with it, you'll be the best team ever. When you don't, start treating it, ⁓ testing as a circular thing. Right. Where you got these automation people who kind of free up time for other people to look up things. And it doesn't need to be the same person, but you kind of go in the circle and each person just moves that moves it forward. And you're going as a wheel, you're going through your products as a wheel, kind of moving it forward. And, ⁓ that's another way to kind of that I've seen people phrase it. ⁓ and you definitely have your corner of the testing space. ⁓
speaker-0: Maybe maybe I do. And again, I th and I think ⁓ Mara's one of those other persons that I've been following for a while. So and definitely ⁓ the answers that I give and everything that hell will definitely ⁓ have been influenced by those people too. ⁓ Yeah d I've not I don't think I've heard that that circular kind of thing before, but for me ⁓ there's a lot of value in putting people together in Two people, one keyboard, basically. Yeah. And I know that there's loads of people who love it, and there's unfortunately many contexts, but it's frowned upon because it looks inefficient, but it's not. ⁓ there's a lot of value in there. And and again that it so the the the the person or the group of people who performs a specific kind of task, ⁓ that no matter if it's development or testing. specification or or the the ops work. who actually is gonna do that highly depends on the context and and and team composition and individual skills and and team cohesion and and communication and all these kinds of wonderful things.
speaker-1: Yeah, I think team cohesion ⁓ solves a lot of problems that like other technical things people try doing. just like, if everyone trusts each other, word you used earlier, right? Like, and users trust each other, then it can do more. ⁓ And I think kind of that also leads to people needing to be a bit more comfortable in areas that they're not necessarily more comfortable in. So in the past that
used to be exploratory testers being told to read code, for example, ⁓ or ⁓ developers needing to write tests. I think is the other one where people start doing that. ⁓ How do you help those people build the technical confidence in either direction from a exploratory tester going into automation, developer going you need to write more tests now because it is actually your code. You need to prove the quality of your code. How do you train those people?
speaker-0: So I definitely have a lot more experience with the former with ⁓ Teaching testers and and helping testers take their first steps on that automation journey. ⁓ I have worked with developers before. ⁓ so let me get back to that because there's one specific thing that I want to highlight just there in a moment. But for the ⁓ Teaching, mentoring, whatever you want to call it, these ⁓ testers on their ⁓ first ⁓ to take help them take the first steps in automation. so what I've been lucky enough to do to be able, be allowed to do several times now is just work with a small group of people for a number of hours a week and just pair program with them. And That typically starts with ⁓ me just typing a ⁓ creating a very simple test or or or ⁓ maybe a step before is that setting up their development environment. Because that that's where things typically start breaking down with people who don't have have just never done that before. And and that's absolutely ⁓ that's ⁓ obvious that that's normal because you've never really done that before. And it's just ⁓ one thing that I I think I am kind of good at is ⁓ breaking these seemingly complex things down into and and allowing people to take very small steps. And then it it's just as setting up a development environment and writing a first test and running that test. And ⁓ Putting that on a version control and making that part of a build pipeline, either an existing one, or we set up a dedicated one just for running that test. And sure, at first it will be ⁓ just watch me do it and then repeat after me. And through but Very soon, and and sometimes that's within the first session, and sometimes that's after two, three sessions, maybe. people start to get confidence. ⁓ I can actually do this. And what happens if I change this? And what happens if I add this? And that can be as simple as just one method call or one line of code or whatever. ⁓ but it's that initial feedback of ooh I can actually do this and it works on my machine too. That is ⁓ that's often the hardest once you've got that momentum people start to take slightly ⁓ bigger bigger steps and then it it's not much ⁓ not so much a case of ⁓ giving people the confidence to take the next step but slowing them down. And and making their steps smaller again and say, well, okay, I see you do this now. What are you doing and why are you doing this? And not just ⁓ okay, yeah, let's write 15 more tests or fifty more tests or yeah, whatever. And then ⁓ but at some point and that that is fairly quickly typically, because people are smart, ⁓ in general, and ⁓ I you you you almost visibly see that light bulb go on above their heads. So ⁓ I can actually do this. And and and again the trick is to just talk them through it, walk them through it, make it them ⁓ make them do it themselves, take s and take very, very small steps and and ⁓ because if if I asked you out of the blue to just build a car, that's gonna be incredibly difficult. So why not start by filling up the patrol. Just add it's a f a very small and and and ⁓ take a very take very small step, change a tire maybe. those kinds of things. And and just take small very small steps and the feedback and and have people say, well ⁓ this is what I'm doing now and this is why I'm doing what I'm doing. And ⁓ sooner rather than later they will they will run with it. That's absolutely it's it's a Privileged to see you really.
speaker-1: That sounds awesome. That must be a real great thing to see how people kind of move through those processes. So I want to pivot slightly. ⁓ You're in automation. I'm in automation. I want to do a bit of talking around tooling about some of the things you've ⁓ I'm sure you've seen as many horrors as I have. in the testing space where everything had to be an end-to-end UI test. But you've been a massive proponent for shifting to more balanced testing, so API integration, contract testing. ⁓ Why do you think so many teams still struggle to step away from that safety blanket of the UI test?
speaker-0: I'm I'm not sure it's it's necessarily just a safety blanket. Although there there's definitely ⁓ a a part of it is because well ⁓ that's the only way I can really know that my system as a whole behaves in the way it's intended to behave. but ⁓ generally and and if I look back to how I started as a test automation engineer, well I was literally just given a stack of Excel sheets. Very a lot of very long Excel sheets and say, Well, you go automate this. And all those that what they were regression test cases. They are they are written by humans for humans. And at some point the humans got bored with it, and then they said, Well, let's go automate this so we don't have to do it anymore. ⁓ but they are written from the perspective in which a human interacts with a piece of software at the end. user perspective, which heavily involves that one interface in your application that's explicitly written to be consumed by a human being, the graphical user interface. So I think a lot of it comes from that, and and that's how I started 20 years ago. And it's unfortunately so I'm I'm working with the consultant client and another that's struggling with the with that exact problem still. So that that's sometimes on the one hand, that's sometimes frustrating that it's it's well 20 years down the line and we're still trying to solve the same kind of problems that we did 20 years ago. ⁓ the good thing is, well, it it pays for my mortgage. No, but so it is a lot of it is People only believe
what they see and and these UIs and these N2N as they're highly visible. They do, ⁓ that's how the system works in production too. And sure, there is value in a small, good, reliable set of those N2N tests. but again, and I keep coming back to well, what it is that ⁓ what is it that you want to know about your system and what's the most efficient way to get that information? Yeah. And that might be an end to end test, but in a lot of cases it doesn't necessarily have to be an end to end. It's also why I and I don't want this conversation to be a rant around pyramids. but I I try and and avoid talking about that too much these days because we're focusing so much on end to end or integration or unit. So ⁓ first of all, I don't know what those words mean anymore. I agree. so again that's why I I the only thing I really what I'm really interested in is ⁓ well literally valuable feedback fast. It it's ⁓ information about this the state of the behavior of my system. Do I know what I need to know? And then ⁓ what's the most efficient way to get that information? And that's sure. a big part of it is ⁓ breaking down those end to end tests and and decomposing them into API level, code level, contract level kinds of tests, but ⁓ by not framing it as ⁓ we're moving from end to end to integration or to to unit or to component or to whatever kind of test, but ⁓ just so this is what I need to know here and Let's see what we can do to find that information in the most efficient way possible. That helps me a lot, at least.
speaker-1: Do you, ⁓ did you ever read Google's testing on the toilet series that came out like 15 plus years ago?
speaker-0: Absolutely.
speaker-1: One of the things that I really liked and I and I agree with your names don't really matter any at this point ⁓ is that like It must have been Let's see. We're in 2026. Yeah, it must have been coming out in about 2008 that series and ⁓ One of the ones that really resonated with me was stop calling tests integration tests unit tests and to end tests. Small, medium and large is how they described it. So an end to end test would be large, but also if you have a huge API test that's kind of starting up the world, that would be a large too. Right. ⁓ And I like if you got your integration tests, ideally those should be the mediums and then your ⁓ unit tests. Those should be the small tests, but like it's more how quickly do they run and how much of the world do they start up just to run? ⁓ And that was the one that always really, really resonated with me. And it's something that we still use quite heavily in the Selenium project, in our build system when we're building and testing is that we can just go run these tests, but only run the small ones. And it's like, it runs. Cause that way you get different types of feedback. yeah, the pyramid, if you were just using the old terms of kind of units, integration end to end, they go a bit haywire in my mind. Richard Bradshaw calls it the testing hourglass because it suddenly becomes inverted at the top where it's like two pyramids on top of each other. I call it the testing Christmas tree. because like it goes up and then out and up again, like depending on how people seem to do things, it's kind of weird. So I totally agree with what you're saying in that. And ⁓ to me, it's still down to the maturity. you'll be to people listening, you'll be surprised you can walk into a bank who are still doing things exactly like they were doing nearly 30 years ago.
speaker-0: I I I I was talking to someone about this last week. I think I don't know exactly where and when, but we all have these visions about these big companies that are incredibly successful in what they do and they make a lot of money doing what they do, and we think ⁓ those people must have really figured it out. But ⁓ they don't I am I every time I and and that's even with ⁓ the the the the the the it it doesn't really have to be ⁓ an incredibly big company. It can also be so at a super successful scale ups or whatever.
speaker-1: ⁓ yeah.
speaker-0: ⁓ no, those people that those web first, mobile for internet first kind of companies. So ⁓ they must have figured everything out and they must be doing things perfectly. Well ⁓ as you y you've per you've talked to so many people and teams too, and I have had the privilege of doing the same and it's It's the s you see the same weird let let let's be nice and say weird and interesting things everywhere. Absolutely.
speaker-1: Yeah. And then you have teams within companies who are way further in the maturity model than other teams within the same company. And you get all these weird. I remember I was quite shocked. I moved from a startup to Mozilla in about 2010 and then spent that next decade working in Mozilla. And I remember when I joined, I was shocked to kind of the startup I was at.
speaker-0: ⁓ absolutely.
speaker-1: some of the CI processes and things like that were leaps and bounds ahead of what Mozilla was doing. I'm like, but like, you are the web, you are doing everything so fast and like setting direction and being open source and amazing. Yet some of your practices are like, why?
speaker-0: ⁓ the this is so familiar, yeah. And again, it if you ⁓ the the the the benefit of of the one of the wonderful things of the what I do is that you get to see so many different teams and companies. So ⁓ but it's it's and that it it's this it's the pretty much the same situation everywhere, as you said, that there's they they do things and and we all know or we pretend to know like how it should be done. It's just that it's not often that we're actually doing it. And then maybe some teams within a larger company are doing things or at least doing things better than than others. Yeah. There's we've still got our work cut out for us.
speaker-1: ⁓ yeah, definitely. And I think this leads really nicely into kind of the next section I wanted to bring up was around like fostering skills and ⁓ using AI to
kind of move us along the maturity model a bit. ⁓ One of the things that you mentioned earlier is that you favor pair programming or ensemble programming, however you want to call it. ⁓ And that's a, to me, I I've always been a big fan of that because it's very collaborative. ⁓ and I think it fits in nicely with the way humans tend to work. ⁓ because we're social creatures, ⁓ as much as the kind of the, the average engineer likes to be an introvert and things like that. At the same time, we're still somewhat social, ⁓ creatures. how do you see kind of the move with AI at the moment? Cause to me. using an LLM is kind of like pair programming. If you're doing it, ⁓ where you'll hate the term, but it's the only time I can think of at the moment, human in the loop, where you're kind of asking questions, you're talking. How do you kind of see that in your kind of work, but also how is that, is that changing the way you're lecturing to people at the moment?
speaker-0: honestly not really. ⁓ but that's mostly because I've always focused ⁓ not so much on specific tools or techniques, but very much about the fundamentals of the things that we do. And and I honestly believe that as long as we're ⁓ building products for human beings, that it's good to have a well, as you said, the human in the loop or that and how however much I too despise that term. But it's a it's a great assistant and and a lot of it it depends on how much of your autonomy or accountability d are you willing to hand over to assistant. I d I was literally yesterday. I was ⁓ with a consulting client. ⁓ together with ⁓ a few other consultants, we we ⁓ we support a range of development teams ⁓ to help them do their test automation more efficiently and we do some proofs of concept ⁓ with them and we work and so and and ⁓ the the the person sitting there said, Well yes, ⁓ I've been ⁓ burning tokens like a madman and writing all these tests that absolute look absolutely wonderful. I just don't have a clue what they're actually doing anymore. But they look good. So that's the ⁓ that's a problem that I see. That's a chance and I know ⁓ that that that specific person he's been in this industry for a while and he he'll take a step back. two steps back and say, well what's it that actu that I'm actually doing? But ⁓ I'm Scared is maybe not the right word, but it it's I see a big risk in so many people and teams handing over a lot of that knowledge and the thinking and the accountability to well not the accountability because we are still accountable, but yeah, a lot of the the responsibility, but also the responsibility of learning and understanding and learning to understand the system to those LLMs. Switch ⁓ I hope it's not gonna backfire too hard, but I'm afraid it will at some point where
speaker-1: Yeah.
speaker-0: How how are the the people who so I've been we've been in this field for twenty years, so we've we've seen that and and we know that world before LLMs gave us all the answers and maybe some of the right ones too. but how are the people who are just entering this field going to get that same kind of learning and that experience building? Because ⁓ in all honesty, I don't think The human brain is really made for the pace at which you're required to learn and understand.
speaker-1: I think that is quite correct. ⁓ One of the things that I've noticed in kind of teams and speaking to people is it feels like people are heading towards burnout a lot quicker than they used to be because they had the time to think through problems, solve the problem, push it out, then wait for the feedback and have those loops and kind of In that you have these built in five minute, 10 minute breaks while you're waiting for a compile, waiting for a CI, things like that. Like that would just naturally there. And now it's like, I finished this. If I do a left arrow, I'm into the top level of my, ⁓ my master agent. And I can then view what other sub agents have finished and I can go look there. Like they don't have that anymore.
speaker-0: And and do we need that. Our brain needs that. It needs that time that to let things sink in, to digest things, to ask questions, to form those neural those neurons, those neural connections to actually learn things. and and I'm curious to see and quite skeptical about how that's gonna turn out in a couple of years, in all honesty.
speaker-1: Yeah, think, yeah, Simon should put a blog post out. He's on ⁓ rocket powered jetpack, ⁓ jetpants.com, I think is his blog. But ⁓ he put out there that like one of the biggest concerns he has is ⁓ the lack of juniors having innate knowledge in what they're doing, right? They solve the problem because LLM, they've tested it, it works. They push it. But they're not understanding the who, what, where and how of kind of the code that
speaker-0: No, that that's exactly what I'm referring to. And and another dimension to that is and I've I've talked about that rid about that before, is LLMs are great at doing ⁓ more and faster, but ⁓ we also need to make things better and and and truly better, and that requires investigation and thinking and ⁓ discussions and trying things out and so ⁓ who's gonna do that? I I'm not sure. I'm I'm I'm again it it is I'm starting to sound like a lotite here. ⁓ which it's not 100% untrue. Let me give you that. But it is let let's let's put it onto professional skepticism. it is ⁓ the the the so the the I've got a a a master's degree in computer science. So that part of me thinks it's absolutely fascinating. ⁓ that the technology and what we can do. but the quality part the human and the learning part of me is very, very skeptical about all of this. And especially in the longer term. And I know there's not a lot of companies that are really good at looking at the long term kind of things because that's that that they've ⁓ that that was true before AI and LLMs by the way, but it it it it doesn't help. It doesn't help really.
speaker-1: Well. So what's
that one foundational skill or mindset that you think people need to focus on over the next few years to kind of protect themselves, their career, things like that in this new age?
speaker-0: That is a good one. That is a good question. So for me, that is probably that it's definitely not ⁓ writing better prompts but no it is ⁓ all jokes I think ⁓ critical thinking and not taking what's being thrown at you by an LLM ⁓ or by your coworker or by your manager at face value and The ability to ask the right questions and to Communicate talk about find and talk about problems and and find ⁓ ask the questions about all the ways in which things might not work. ⁓ so in in in a sense I think that the skills we've always had or at least that we've always needed as testers. They've they're only becoming more more important as the whole part of writing the code and writing the tests is increasingly often being done for us and increasingly at at a breakneck speed these days.
speaker-1: Yeah, cool. Right, Bass, we've reached the point where we are it's time for our quick fire round. This is where I ask short, sharp questions. Your answer needs to be very succinct. ⁓ But it kind of gives me and people watching an opportunity to kind of see how your mind thinks on other things. So hopefully, like they get to see another side of you. Are you ready?
speaker-0: I am ready, absolutely.
speaker-1: ⁓ What is your favorite testing or development tool that you've used recently?
speaker-0: ⁓ mutation as a technique, mutation testing as a tool, I absolutely love striker.net at the moment.
speaker-1: What is a book that every QA lead or software engineer should read?
speaker-0: so business books, ⁓ well you can't go around taking this thing seriously by James Bach, Michael Bolton. for the I'm a big fiction reader, so ⁓ I absolutely love Razorblade Tears, or basically anything that ⁓ S.A. Cosby has ever written.
speaker-1: Awesome. What is the most underhyped testing metric that teams just ignore?
speaker-0: Mutation coverage.
speaker-1: Just for people, what is a mutation coverage? one line.
speaker-0: So that is ⁓ the percentage of mutants generated by a mutation testing tool that are actually killed by your test switch.
speaker-1: Cool. What is the most underrated skill for any technical tester today?
speaker-0: Thinking like a tester and not just like a developer. Thinking about valuable feedback and not just fast feedback.
speaker-1: Perfect. ⁓ If you weren't a test automation consultant and trainer, what do you think you would be doing now?
speaker-0: I don't know. I would absolutely love to restore Classic R, so my dad's been an auto mechanic for over 40 years. ⁓ but I definitely didn't inherit his DIY skills. So it's probably best that I stick to this, to the one I'm doing right now.
speaker-1: Amazing. ⁓ Bas, this has been incredibly awesome, practical, insightful conversation. ⁓ Thank you so much for taking your time to share your expertise with our audience today. ⁓ Where can people follow you? And do you have any parting words to add?
speaker-0: ⁓ parting words not so much. I think I think ⁓ just apart from thank you for having me and to those of you who are listening, thank you for listening. I really appreciate it. if you want to get in touch, ⁓ my website is ⁓ on testautomation.com. ⁓ you'll find links to my email there and to my newsletter. I on and off LinkedIn depending on when you ask. I these days I try to not spend as much time as I used to dare. So ⁓ email and and just my website is is a great way to to get in touch and to keep in ⁓ to keep up with ⁓ what I'm up to.
speaker-1: Amazing. Thank you so much for your time today, Bas. This has been really insightful. ⁓ And to the audience, as always, make sure you hit the Subscribe button on Browser Stack Talks to catch our next episode. ⁓ Until the next time, keep testing and keep chasing valuable user feedback. Thanks for tuning in, everyone.
speaker-0: Thank you.
speaker-1: All right. That was awesome.