The last year in product design has felt turbulent. Not just because the tools changed — but because of the sheer volume of extremely confident opinions about what the future looked like. Design is dead or whatever bullshit.
You’ve been on LinkedIn, so you know.
For me the volume of that certainty has been the exhausting part, and has manufactured an anxiety of “am I approaching this the wrong way?” in me — a normally not-anxious person. Because unlike the periods of change I’ve worked through in the past, getting this one directionally wrong has felt expensive — not least because this particular era has coincided with some aggressive economic circumstances.
All of which to say that I spend my days with designers asking me whether making the wrong moves with AI might cost them their job. That’s real and that's a heavy burden we all carry now.
So Config was unexpectedly reassuring for me. Not because it provided many answers, but because it revealed that almost everyone is converging on the same solutions. Even more reassuring was that attendees and speakers alike all sounded remarkably similar to the teams I work with — wrestling with the same questions, experimenting with the same workflows, feeling the same excitement and the same discomfort.
Best of all was was hearing that the ways the world’s most aspirational teams have adapted consist of things my designers and I have done, or are actively doing. Designers writing code, rapid prototypes replacing static artifacts, smaller bets shipped earlier, less time in the pixels and more time shaping intent etc. More than once I left a workshop with the feeling that while I hadn’t learned anything new, I was reassured by the idea that we were all moving in the right direction. Hearing things like “you need to set up your design systems for LLM legibility” didn’t feel like an attack — I actually understand what needs to be done, and why that work is important.
I also noticed how almost nobody positions AI as some incumbent force that would soon arrive and fundamentally change everything. Instead AI just felt normal… embedded, assumed, almost boring. I recall thinking a while back how we’ll soon be thinking about “AI Powered” the same way we thought about databases, or AJAX back in the day — just not all that interesting.
That was the vibe.
We’ve moved from “is AI useful?” in 2024, through “how do I make AI work for me?” in 2025, to “AI is the capability unlock that allows me to create the work I’ve always wanted to create.” in 2026 — which is profoundly different framing and fundamentally more exciting.
There were three themes that kept resurfacing throughout the week — in talks, in workshops, in hallway conversations and then while staring at the seat back on the 14 hour ride home:
Figma announced ↗ a lot of new functionality that pulls code closer to the design process, which feels both inevitable and necessary if they want to remain relevant in the long term. Kinda interesting, but also could’ve just watched the 2x version at home. But what I did find interesting was how they repeatedly referred to code as a material. I like that framing because it’s familiar — designers already work with materials. Pixels are a material… words are a material… motion is a material… systems are a material. We use them as tools in service of an outcome.
For years the industry has been trapped in a fairly unhelpful argument around “should designers code?”. It has always felt a little binary, as if there are only two possible futures: designers who remain entirely visual, or designers who inevitably become second-rate engineers. So the conversations at Config felt like a clarification to this, almost toward ambivalence. Nobody cares if you code. Instead there’s an expectation that if something exists in Figma, it exists in production too — and that instead of being CSS magicians or whatever, designers are responsible for narrowing the gap between concept and code. Now that feels less like a philosophical shift and more like a practical one.
I see this code-as-a-material shift as similar to how understanding color theory, typography or art history are useful, but they’re not required for designers to efficiently and accurately solve problems. In the way that I don’t need to understand the mathematics behind a mesh gradient to use one, now I also don’t need to understand the precise structure of a React component to create one for a feature either. A means to an end.
This all feels like justification for me dropping out of React Bootcamp in 2024, but the idea was underpinned by two recurring themes:
One thing that came through repeatedly in talks and workshops was the idea of ownership. The strongest teams weren’t using AI to push designers further away from implementation, nor were they creating more process around the existing design and engineering divide. Instead they were building environments where designers could safely own more of the customer experience — in whatever form.
That naturally creates a need for designers to operate comfortably in (abstracted) code, but importantly, protect them from the complexity inherent in modern engineering. The expectation wasn’t that designers become full-stack engineers, it was the importance of building confidence to make production changes, improve interactions, challenge implementation details and explore ideas that might previously have died during scoping or handoff.
The companies doing this well seemed remarkably consistent in how they approached it. Designers had local environments, they operated in PRs, wrote specifications, they were encouraged to make mistakes, ask naive questions and slowly build technical confidence over time without threat. I resonated hard with how Frederick Andersen ↗ spoke about how designers have perfectionist tendencies and imposter syndrome — so the emotional cost of looking “dumb” is much higher than we might like to admit.
Ultimately producing coded artifacts no longer feels optional in the same way that communicating well via Slack or contributing to the design system isn’t optional. It’s simply becoming part of participation in modern product development. The discomfort is there, but it’s a discomfort worth leaning into.
I found another interesting observation hiding beneath a lot of talk topics — that AI increasingly behaves less like software and more like a new colleague. And like a colleague, context determines whether the relationship succeeds or fails.
This came up repeatedly, but it’s probably worth saying directly for the people who don’t particularly want to hear it: for this reason, the era of “I’m just a bit messy in Figma” is coming to an end 🌶️
Unnamed layers, detached components, duplicate tokens, ambiguous file names and infinite canvases of versions will prevent the systems that are helping us design and ship faster from doing their job. If we want AI to understand our intentions, we need to make those intentions legible — and we need to be structuring our files, systems and components in such a way that allows for flexibility but also conformity #sorrynotsorry
Agents in Figma, Claude Code and whatever comes next simply cannot operate effectively without that context — they need names, conventions, specifications, version history and a shared language. And you need AI in your files to be effective… so while we might not like it, these are the terms. This was clear listening to the workflows of teams like Figma, Waymo, Microsoft etc — they weren’t the ones who were better at prompting… they’re the teams with the clearest systems and the most consistent and legible patterns.
That’s a very different competitive advantage than I expected to hear — but a lot more actionable than the assumption that speed is the only differentiator. All this comes from some variation on the idea that humans are remarkably good at figuring out intent and meaning from incomplete information… AI is surprisingly bad at it.
The second theme that kept resurfacing throughout the week was the inherent risk of speed.
AI allows us to move incredibly quickly, which also allows us to move incredibly quickly in the wrong direction. The incredible Catt Small ↗ from Dropbox spoke a few times about this tension, and how they mitigate it by assigning more trivial work to AI while reserving thought work for humans. I understand the logic, but I suspect that distinction becomes increasingly temporary as the models improve. Another speaker elaborated on the sentiment, encouraging teams to allow AI to explore aggressively in spaces where decisions can be easily rolled back.
Both resonated with me because what I was hearing wasn’t really speed itself — more the risk of certainty. Or more specifically, the speed at which confidently-incorrect assumptions harden into decisions. For me, the recurring theme across the customer research, validation and collaboration-themed talks was the importance of building conviction rather than simply building solutions.
One speaker asked “Why do I want this to be true?” in reference to blindly agreeing with AI — which stuck. In ‘From concept to confidence with Dscout AI’ ↗ they spoke about how speed itself is no longer the differentiator. Instead — knowing what you stand for, what you won’t do and building conviction is becoming more commercially important than ever — and perhaps more difficult now that we’ve democratized execution and decision making.
Bobby Meixner from UserTesting ↗ presented research data that made the emotional side of this more tangible. In a recent study they found a 6:1 gap between how fast designers are moving with how confident they feel — 91% said their work is moving faster with AI, but only 15% feel more confident in their output. Tons of other amazing stats which I couldn’t write down because each slide lasted 5 seconds — but one of those “thank god it’s not just me” moments.
There was also a sentiment that while AI compresses execution, it doesn’t compress experience. So in a nice industry full-circle moment, the discussed solutions seemed to lie in customer conversations, research, critique and better collaboration between designers and their broader teams. The ability to explain not only what we want to be true, but why we want it to be true.
Ultimately I liked that moving fast and moving thoughtfully are not opposing forces — they’re simply two separate races that need to be run at the same time. For me, this became one of the clearest distinctions between companies using AI as an unlock and companies recklessly chasing profit speed.
Actually, this turned out to be one of the most reassuring “ah-ha!” moments for me — that the companies that felt genuinely ahead weren’t the ones celebrating replacing five people with one person or token-maxing their way to higher throughput. They were the companies excited about being wrong twice as fast, learning faster, building conviction earlier and executing with greater confidence.
The final theme was the importance of craft — but not in the 🐸 “graphic design is my passion” kind of way.
Instead there were a number of talks centered around building experiences that were differentiated by feel, and by an emphasis on the human experience itself. One speaker, who I failed to attribute in my — quite frankly — awful notes, asked: “What do we want users to feel? What makes them come back because they want to?” — which felt like it summed up the real design resistance.
The idea that 10,000 of us gave up good coffee and squeezed into a convention centre because, not-so-secretly, we all just want to build products that people love to use. Instead of becoming better prompt-people.
Then in the closing session, musician RAC spoke on the topic ↗ ‘It used to be hard to be average’ — about the immense gravitational pull of sameness becoming the biggest threat to creativity. In contrast to the polite applause Dylan Field received in his opening keynote, this was the topic that had the crowd actually roaring.
RAC wasn’t alone on the topic. Rose McManus ↗ from Instagram spoke about the strangely human identity crisis that emerged when people were allowed to choose their own usernames. Ryan Powell ↗ from Waymo talked about how designing audio notifications in the key of C made riders feel more reassured when handing their safety to a self-driving Mercedes. Matthew Ström-Awn ↗ spoke about how attempting to remove whitespace in Stripe led him to overthink pretty much everything. All of them were talking about the same thing… how taking something average and injecting tiny, human details can have a disproportionate difference to how it feels.
There was probably some compelling financial reward to all of this stuff, but conveniently the economics weren’t mentioned once. Which definitely felt uncomfortable in a room full of people who are now required to justify their existence to corporate cultures demonically-focussed on profit. But I digress.
Beyond making cool things, there was a consistent undertone around the importance of simplicity in a world of increasingly complex, but ultimately unhelpful products. Jésabel DC ↗ referenced principles behind Lego instructions: “If it needs an explanation, it needs a redesign”. Which feels like a personal attack increasingly relevant.
The speakers weren’t really talking about craft as a matter of process or aesthetics — they were talking about craft as a source of humanity. As our ability to create anything becomes increasingly abundant, the value shifts towards deciding what deserves to exist in the first place. What do people need to feel… what earns trust… what creates delight… what makes something feel considered rather than merely functional.
I think that’s was people were reacting to in RAC’s talk. It used to be hard to be average, and now average is free. Which “raises the floor” or whatever people say in these moments — but practically it means originality, empathy and judgment are becoming the things that matter more than ever. And that the more capable our tools become, the more valuable our humanity becomes alongside them.
So yeah, worth the brutal flight.
I arrived at Config hoping for answers and mostly left with reassurance instead. Reassurance that the things we find difficult aren’t signs that we’re behind. Reassurance that the uncomfortable parts of this transition are uncomfortable for everyone. But also reassurance that pushing through this discomfort is pushing towards a future of more ownership and capability… more humanity and more ambition… and that’s a future I have a renewed excitement for helping build.
Jess from Everyday UX has some really keen insights from the event worth reading. And Joey from Baseline does a really thorough technical overview.