Ep. #13, Ten Times Faster at What? with Jennifer Riggins

Heidi Waterhouse: Welcome, Jennifer Riggins. Would you like to introduce yourself?

Jennifer Riggins: Sure. I'm Jennifer Riggins. I brand myself as everyone's favorite technology journalist and host. I've been doing this for about 15 years as a freelancer for different media and different media types too, not just publications. And I specialize in the developer experience and that more impacts developer productivity.

And that was platform engineering until two years ago. And we can all guess which two letters is the majority of what I do now.

Heidi: That is fair. When we thought about bringing you on, we were having a discussion in the background and we're like, oh yeah, definitely Jennifer. She's the OG progressive delivery person. So we're very excited to have you on.

First, I'd like to start with, what do you think Progressive Delivery is? Did we name it right? Because you were there when it was born.

Jennifer: I mean, James Governor gave birth, but I saw the baby soon afterwards to the name. Still love that. Would love to coin a phrase one day.

No, I think it's good. It's interesting because certain aspects of progressive delivery, like chaos engineering, we now recognize was not a good term for the majority of people because enterprises got squirrely about it and it was very nervous.

It's a great practice. It's a part of Progressive Delivery, but I would say Progressive Delivery makes sense. It's progressively delivering or receding delivery, taking it back to reduce blast radius. So I think it makes sense and I think it's a safer way to go about it.

I don't know that most companies are progressive delivering AI right now. It kind of feels like we're ignoring the best practices a little bit. All of them: small batch size, have guardrails in place, everything.

But all of these progressive delivery best practices and DevOps best practices that were good before are more important now.

But we just beat our heads against walls sometimes. And that's okay.

Heidi: What do you think AI would look like if it were progressively delivered?

Jennifer: Well, I saw a very interesting talk today at API Days or whatever we're calling it now, Future of Software Technology. I don't know, FOST, they're calling it, but API Days, basically London, by, I'm going to bastardize his name, but Wynand Vorster, and he was talking about agentic control and everything he was talking about, besides maybe we need to rework the org chart, which people haven't really talked as much about as they should in job descriptions.

But the rest of it, this idea of coming from pilot to production, is shocking that we're not doing it more. I think at well-regulated companies, we are doing pilot to production, but at most companies, I feel like they just hit the ground running without the guardrails in place and are going to start falling off the cliff because they didn't do those steps to create those boundaries, to be intentional about data, to be intentional about all the Progressive Delivery and DevOps practices that, again, we've been talking about for 10 plus years.

Heidi: Yeah, I think it's one of the interesting things that we struggle with is, very little of this is actually new and it's very much, let's talk about people and process and not as much about the tools. You need some tools to make it work, but they're not new tools. We've had these tools. So I think when you talk about, why are we not using guardrails?

I think we don't understand where the road even is. And so we don't know where to put the guardrails.

It's like you don't put guardrails on big flat sections because nothing bad's going to happen, but you do put it on curvy mountain roads.

And we don't know what the curvy mountain roads of AI are right now.

Jennifer: Yeah, that's true. Or if they're on a different planet.

Adam Zimman: Or one could argue that it's all curvy mountain roads.

But I think that, to your point, Jennifer, we're seeing this consistent aspect of teams that actually were really good at process and automation before we called it AI are proving to be way more successful than the organizations that were good at that before we were asking the robots to automate all the things.

And it turns out that when you automate chaos, chaos ensues just faster.

Jennifer: And I think it's telling. We don't know what happened in the background, at least on the record, of why Google dropped DORA. But considering the last results were not that exciting. There were always exciting things around, not boring, but not sexy stuff, like documentation and explaining stuff. But writing code was never the bottleneck and never the interesting use case of anyone's job for the majority of their job.

So I think it's very telling when a huge provider of AI and a huge provider of everything, basically, that's backing it, was doing this wonderful research project. And then they started diving that research project, which everyone was really on board for. And then they got it about AI and it started not looking useful at all. So they just stopped the project. Will the people continue on? Will the research continue on? Yes, but probably not as transparently and at that scale because it does not have that sponsorship. But I think that just says something really concerning about gone all in on AI, but it's not great. But we're still going that way.

Heidi: I think that has been such a foundational report. So many other pieces of software have their reporting metrics built around it. Whether or not that's a good idea is definitely debatable, but it's real. And to pull it from the community is kind of a strong choice.

And it does make you wonder if the feedback that we're getting around, in order for your team to be productive, you have to have a high-performing team that has to feel psychologically safe, if that is a message that the powers that be aren't excited about anymore.

Jennifer: This also happened when the doctors released their reports about AI ethics, and Google got rid of that, too. So that's dark.

Heidi: It's a little bit dark, I think.

So jaywalking didn't used to be a thing. It didn't used to be a crime. People belonged on streets and everybody else moved around pedestrians. And we only instituted the idea of jaywalking after vehicular traffic really came about.

And so I think in a lot of ways, we are blaming the victim, the pedestrian, for being in the way of the fast thing. Well, if you get hit, it's your fault for being in the way of the fast thing. I don't think that necessarily makes it right, but I also don't think we're here to just hate on AI.

It's doing a lot of interesting things. We're doing interesting things with it. And one of the things that I think is most interesting is, where do we go if code is not the bottleneck? Have you seen any movement around where that's going to be?

Jennifer: Well, it's interesting because we're also deploying agents, but in the same systems that weren't working before. So again, we know there's companies, there's enterprises doing it well. And I had a really interesting conversation with Lloyds and same with NatWest, seemed to be doing really interesting things.

But they're not going all in on AI.

They're going all in on the boundaries around AI and breaking down data silos before they go in on AI.

So, which we always knew, the problem was never getting robots to replace jobs. It's the same problem we've had for the last couple years.

The tech industry, not accidental tech companies like banks that seem to value their employees better, but the tech industry, the last couple years, decided that they're not responsible for their employees. And so they're looking for shortcuts to replace their employees that cost an astounding amount of money to hire in the first place. And then they're just like, oh, we're giving up on this ethos. But the companies that invest in their employees and invest in their employees' tools that help foster their success, safety, security, are still doing well. But companies that don't care, they're the same companies that blindly laid off people and are pushing, shoving AI down people's something.

Heidi: You said something interesting about data boundaries. Can you go into that a little bit more? Is that proprietary?

Jennifer: No, we've always known that the benefit of AI would come from the individual data. I'm sorry, ChatGPT, I don't even use it because what they did in Kenya, but Gemini is great for helping me lead my conversations with doctors about menopause and with my kids figuring out how old is that tree.

It's not that interesting for a business because it's generalist. And if you're a 75 or 150 year old bank, though I think again, you'd probably want to wait on the last 10 years of data.

You're interesting. I always say, I don't use it to write because that's my unique value proposition, is my writing and asking questions and having conversations with people. Do I use it sometimes to explain to me a new tech concept?

Maybe I'll use it to research, but it's not great at that for me. It's good for advocating for my healthcare. It's great. And I've also given up on my own personal privacy.

But this idea that a company would want to use a public Copilot or a public anything would make no sense because the most benefit it has is its data and its unique value proposition of whatever it offers. So it should not be using AI for that.

So if you believe that your unique value proposition is coding, then you shouldn't be using it for that. But it seems that nobody does.

But the man I was talking about earlier from API Days, he was talking about, our whole systems are wrong. And granted, he works with private equity company advising people. But so money is involved. That is kind of the point of most businesses, we must admit.

And he's like, everything's organized in tech around doing something, achieving something, but not around driving value towards the end goal, which is, let's be honest, profit.

So he was advocating for a reverse Conway's Law: invest in architecture to create this autonomy, but not grouping your architecture around purpose, like your purpose is to test. No, your purpose is to get software that is secure or quality that's not going to ruin your reputation and not lose customers or attract new customers. So you need to restructure everything around that. I found it very interesting.

Heidi: That is super interesting. And a thing that I've heard other places. It's like, you need to back up and stop thinking of your job and start thinking of your value. And I think that that's not something we teach people a lot. It hasn't previously been something we were really encouraged to consider.

Adam: Although on that note, I think there's also the needed distinction that value at any cost or through any means necessary is probably not what we want to incentivize. And I think this goes back to the way in which we see a lot of the frontier models doing things to basically cheat to win.

And that also brings up some major challenges with regards to the outcomes, like saying, hey, let's focus on the outcomes and profitability at any cost. I have no doubt that there's going to be times where the AI completely goes off the rails and does something even more illegal than what we've seen.

And I think where we're also stuck is this idea that accountability is completely lacking at this point. And the humans right now seem to be celebrated for pointing and saying, oh, no, the robot did that. It's like, well, yeah, but I'm not going to delete the program and say, oh, punishment handled. Let's move on.

Heidi: That ancient IBM rule, a computer can never experience consequences. Therefore, a computer must never make a decision.

Kim Harrison: Okay. I've got a related question. Earlier, you hinted at org charts and what they should look like. Let's talk about the people that are involved.

What should an org chart look like? Who are the people that should be involved in leading, building, maintaining, if not only to do it well, but with some, I don't know, ethical standard?

Jennifer: Well, I would hope, but your mileage may vary by company around an ethical standard.

But what this man was talking about was this idea that, first of all, AI does not create 10x developers.

It exposes minus X organizations.

Heidi: Oh.

Jennifer: Because it shows where you have a queuing system, you have this slowness. And again, no one cared that code arrived faster.

People just didn't understand it, which when it comes down to it, it is a problem that C-suite technologists don't speak business at times. And I blame them in part for not advocating for people's jobs because they're responsible for that. But the idea is that instead, and this gets very Team Topologies, you should design org charts around flow, not job title, because especially when it's testing, well, AI is pretty damn good at testing. You need someone more connected to the business and stuff. So I think somewhere along the line, if we're looking at it as don't outsource to AI your unique value proposition, writing tests or docs by itself or writing code was not what adds the thing because then AI can do some of that. It's the people should be your company's unique value proposition. The people you've invested a lot of time in hiring and a lot of money in tech, and then helping them be successful, connecting them to the business goals. So I think there's a lot of senior leadership at fault for never learning that themselves, and then passing the connection of value down, which platform engineering helped a bit.

But now I don't know if we're investing in that anymore.

Heidi: Yeah, that's a really good reflection. And one of the things that I think we want to talk about progressive delivery is how do we see the users of our product feeding back into the product. And I think that includes the systems that we work within to make the product.

It isn't just developer experience, it's company experience.

Because all of that, all of the safety or precarity, feeds into the, do people feel like they are in a place where they can make something good? I honestly think a lot of Google's problem has to do with how they do promotions. They're spending a lot of time trying to decide whether people have succeeded at their promo packet, and it incentivizes new things over maintenance things, and therefore, we get new things and not maintenance, even though the users are attached to the maintenance objects.

Jennifer: Yeah, and I'm starting a book on Spotify actually tomorrow. And the first thing I want to ask is I'm not a Spotify user.

I've written about Spotify a lot. Just an ebook, but I'm not a Spotify user because I just happen to listen to podcasts on my iPhone because that to me is simpler and I pay for music. I don't stream music. So I'm weird.

But I do want to know if they've reached this 10x velocity that they've claimed. Or none of the developers are writing code. We know that them and Atlassian have been saying that since the beginning of 2026. I want to know for the users what has changed.

Because if we do reach this peak velocity, we're going to go too fast soon for the average user. Because people don't like change. And so the end user, or if you're doing it for platform engineering, you're not going to like change.

Are you going to then run a ton of progressive delivery A/B testing, blue-green tests to see which people like more? That would be cool, but I'm wondering if they're doing that or they're just dumping a bunch of new features and they haven't done the research. Kim?

Kim: Big time user. Part of this is because I recently have been living in a more rural area with not great cell coverage. So it's a great way for me to save material, whether that's music, podcasts, ebooks, and then listen to it at will.

They've changed a lot and they continue to offer new features more and more. Sometimes it's a little annoying because you'll open the app and it's like, I'm trying to get back into whatever it is I was listening to. And there's yet another pop-up with a suggestion about this person's going to be near you giving a concert or did you know you could do this with ebooks. They're really extending deep.

You could take a photo of a book that you're reading so that it knows what page you're on and then go to that part of the ebook. There's just so much that they're thinking about and every week there is something new, which could be great if you like it, could be annoying if you just want to get back into chapter whatever.

Heidi: Yeah, see, and I never use it because I want to listen to music in the order that I wanted in, which is frequently a whole album at a time. What if that was the art form was not a single but an album?

Adam: I mean, you can do that on Spotify, but yes, I understand. I'm by no means saying that you should change your workflow to use the app. I do think that it's interesting. I use it not as frequently, but

I use it a fair amount. And one of the things that I feel has been interesting is that I have definitely seen an increase in them pushing the new features and things like that, both in app as well as through other communication channels, whether that's through socials that they know I'm a Spotify user. So I get a lot of ads, as well as emails, they send out emails all the time, even though I say, don't send me any emails. But all of these new things and feature functionalities, it's stuff that I just don't want or need, or at least not from them. And I think that's the weird part for me is that they are definitely trying to grow their platform and all the things that it can do.

And I'm like, okay, I'm not your target user for that. I don't find their platform great for podcasts. I use other platforms for that. I don't find the discoverability on there great. So if I'm trying to find new podcasts, it's not something that I do there.

But I do enjoy it for music, mainly for the sharing function with my kids and stuff like that, to know what they're listening to and things of that nature. But yeah, it's weird how their interpretation of 10x more productive, it hasn't hit me as a user for what I use them for. It hasn't added value to my experience of what I want from them.

Jennifer: It makes me wonder if, first of all, I'm glad they're investing and they're trying on, at least we're seeing things on the other side. So that's good. If you're going 10x, they are producing something that is reaching users.

But I wonder how, and I will ask Samara now, because I'm inspired by that, how psychologically safe it is. Because if they're pushing these new products so much, is it that they're just testing on everyone? Or should they be testing on a certain type of person?

Can you have an innovative canary test where you're putting it on your most creative, certain Gen Z profile user, that you should try on those people who maybe opt in? Or I'm sure they have design partners too that they do this with, that they want the most creative things, or if they're feature dumping on everyone, is that because everyone's job feels like it's tied to the success of whatever feature they make?

Because if it is, let's just be experimental and fun. That's kind of fun. And I'm sure a lot of developers are loving that. And if they're not losing users, they'll reach a point where too many features will piss off users and they'll find that pretty quickly.

I'm sure that their number one metric is probably retention and acquisition, of course. They are a business. But I guess it's just, are you getting inundated with marketing messaging because there's so much pressure for people to adopt these features? Or is it just they want to try it out and they want feedback? So progressive delivery can be very interesting there, but also psychological safety, in this climate, I would be a bit concerned.

Heidi: I'm super excited to read this book. I think it's going to be amazing. Does it have a name?

Jennifer: No, it's been a potential for six months and finally getting the interviews. So we'll see.

Heidi: Excellent. So we can't promote it for you yet.

Jennifer: No, and there's no need. Thank you though. But hearing feedback from actual Spotify users is interesting. Yes, I should dogfood it, but I mean, I also don't code and that's okay to have an outsider's perspective.

Heidi: Yeah. And I think it's interesting because Spotify's Backstage sort of became the platform engineering darling. It's very interesting to see how their rapid adoption can sometimes pull the industry with it because they have enough center of gravity to do that. So it'll be interesting to see.