API Evangelist Conversation with Jennifer Riggins on Developer Experience, Platform Engineering, and the Human Side of AI

Top of mind and heart is always how it affects the developer experience. At an individual level AI is really positively affecting it — people seem to enjoy working with it, and the idea that it lets non-technologists participate in the creation of tech matters, because the future cannot be built only by white men or Southeast Asian men, which has always been the problem in tech and in open source. Nurses should be able to spin up prototypes of what they want, because what gets built then actually serves the people who matter. But at the bottom of my heart I worry about security, guardrails, and best practices falling through the cracks — DORA called AI an amplifier, so it’s a garbage-in, garbage-out situation. The hard part is employment. We’re being told to round off all the corners of our jobs and submit our skills, basically so we can be let go. I worry that we no longer care for our technologists.

Platform engineering is the conduit for successful AI — it’s where the guardrails live, the paths when you want a little more autonomy, and the gates when gates are necessary. It’s proven now as the way of doing AI, and with it comes all the best practices we’ve always had: continuous testing, continuous integration, version control, DevOps, CI/CD, progressive delivery. The thing is, you can’t put everybody under one platform — we learned that the hard way. Data science doesn’t fit the same mold; I heard a wonderful talk years ago about adapting agile for data scientists because it just didn’t work for how their work works. So the silos DevOps was supposed to dissolve, and then platform engineering after it, are still there. The benefit AI promises — being closer to the feedback loops, closer to what customers actually want — only comes when there’s real integration at the API and data level, and we still don’t have that down.

It makes sense for platform and AI to go together, because both of their goals were to eliminate the tedium of the work — first for developers, now expanding to architects and other roles. The AI needs guardrails just like the humans, and you want to make it easier to do the right thing, because both can wander off the golden path — off into the poppy field or get snatched by flying monkeys, distracted from why they were going to see the wizard. But it comes back to the data. We talk about APIs integrating systems, but it’s really about integrating data — that’s the value, the thing people keep calling the new oil. It’s not about the AI, it’s about the data that drives and trains it, otherwise it’s crap in, crap out. A platform is the easiest way to do that translation across all the departments, the mainframe and the cloud native, and to create a common language people can self-serve against with a conversational overlay.

From a platform perspective, so much of what I did was consolidating and making developers’ lives easier — empowering them, guiding them, aggregating resources, providing learning and educational stories. The question now is how we do that externally: how we build more solidarity among all the people in the tech space, because I don’t feel like that’s something we’ve had to do until today. Everyone’s adopting AI and producing more code, but writing code was never the bottleneck — we’ve been saying that for years. We’re behaving in a small-minded way about our code, our technology, our adoption of AI, and it isn’t people-centric. Where’s the joy in the work if you’re just starting a business to eventually replace the humans in it? Especially if we’re going to use AI to replace parts of our jobs, we have to have a human safety net.

There is, and I don’t want to blame AI for it — I think it’s become an excuse. Nine hundred thousand young people in the UK, where I’m based, are unemployed right now; that’s roughly the size of the third-largest city in the country, and the US has a worse unemployment problem and makes it much easier to fire people. The blog I’ve written for the longest just let go of almost their entire staff, and we don’t really know why. I’m connected to a lot of technologists in places like Minneapolis and Chicago that are under siege — people who a couple of years ago were making baller money in tech and are now homeless. I do believe you should hire the best person anywhere for the job, but there have to be jobs. It’s a wider belief that capital is king and labor isn’t — that you get funding, do some magic with AI as a one- or two-person startup, and that will somehow sustain the world. It won’t, and AI is probably the worst thing to try it with.

I’m an advocate for junior developers — how else do we make senior developers? The baby developers coming in now might be truly AI-native, having joined in the last three to five years or even less, and they don’t always understand what they’re talking about, so they end up trusting the AI. You still need to train the new people; there still need to be young developers in the room. To me it’s about bringing the people who know the thing best into the conversation, which might be people who aren’t technical at all, and pairing them with the technology. Pair programming used to be two people; now maybe it’s a business stakeholder, an engineering stakeholder, and AI in harmony. But the human relationships, the people and the process — that’s the socio-technical part, and it’s still the hardest and the most important part of tech.

There’s an expertise problem — Dunning-Kruger is one version, but there’s also the case where you’re an expert in one area and don’t recognize you’re not in another. For something I’m not an expert in, an AI summary can be genuinely helpful — sifting through an overwhelming pile of newsletters, or helping me ask better questions of my doctor in an eight-minute appointment. But I don’t trust a summary of a newsletter I’m an expert in, because I have a unique eye for what’s actually in there. So it’s fine to use AI as one way of consuming information — it becomes dangerous when it becomes the only way anyone is consuming information. It’s a tool. It needs to be part of the school system and part of job interviews. Encourage developers to use it in an interview, because they’ll be using it on the job — why put them at a whiteboard feeling like they’re in trouble at the principal’s office? Put them in a natural circumstance with the tools they’d actually use, and watch how they think.

Smaller conferences. People want to connect, and they still want food — meetups have lived on pizza for years. But the thing I’m seeing demanded more and more, that I can frankly charge more for, is facilitated conversations: maybe a room of fifteen to twenty people around a couple of tables, run under Chatham House rules — what happens in Vegas stays in Vegas, you share the learnings but you don’t attribute anyone’s IP. That’s the wave of the future, because you’re building a relationship and learning an immense amount. Online events I love to host, but you have to offer something unique or people will just watch later. With CubeCon, I’m excited for the co-located platform engineering day and the after-events, not the thirteen-thousand-person hall — who wants to go to a mall and watch people talk? It’s much better to curate a room of people with similar problems or roles, diverse in thought but going through the same things.