Designers aren't being replaced by AI — they're finally being amplified
Surveys show designers are the unhappiest group in tech, but OpenAI's head of design thinks this is the best time in history to be a designer — because nobody has the answers, and people starting today actually have an edge.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
The real reason designers are unhappy is that the role is undefined
In Lenny's tech-worker mood survey, designers and user researchers came out unhappiest on every single dimension: most anxious, most exhausted, least optimistic, and least willing to recommend the profession to others. Ian's read is that the problem isn't AI taking the work — it's that ‘right now nobody can say clearly what a designer is supposed to do.’ Someone trained the traditional way is suddenly told to relearn everything. He also points to a structural gap: engineers have seen 10x or even 100x productivity gains, design teams haven't, because the design process itself is still the dirty work of trying a pile of things, throwing a pile away, showing users, and starting over.
— Ian SilberDesigners doing well share one thing: they feel amplified
Among the happier half of the survey, the strongest correlation was ‘feeling amplified by AI’ — AI making them stronger rather than replacing them. Ian adds to that picture: extremely curious, with time to explore, feeling more creative, highly adaptable. His concrete advice is plain: when you have an idea, throw it straight into an agent like Codex or Claude, let it help you think, help you prototype, try a different way of working. He stresses that ‘something that didn't work yesterday might work today,’ and uses his own example — on paternity leave late last year, coding agents kept getting stuck; a month later, when Codex came out, he tried again and it just worked.
— Ian SilberPM, design and engineering won't merge into one job
Marc Andreessen described the current moment as a three-way standoff between PMs, engineers and designers, each convinced the other two roles will disappear. Ian's position: the boundaries are genuinely blurring, but they won't vanish. He's met PM-flavored designers and polish-minded engineer-flavored designers; a single-function definition never held up anyway, and big companies set levels mainly to slot people into a ladder. But he still thinks someone has to own specific things — a great PM who can pull multiple teams together, narrow the problem space, and manage resources still has value. His advice for startups runs the other way: if you were starting today, hire the best generalists, people who can flow between roles.
— Ian SilberAI is already a great product designer, but not the best visual designer
Asked directly whether AI will become an excellent product designer, Ian's answer is ‘it already is,’ and the key is that it's accessible to every single person. But he immediately draws the line: that doesn't mean it's the best visual designer, or that it best understands information hierarchy, typography, user experience and interaction design. His chain of reasoning: truly good design is designing something that has never existed — the iPhone's multitouch, Instagram catching on that everyone has a phone, the Snapchat interactions that looked strange at the time — none of that had training data. Now it's AI's turn: how to talk to AI by voice, how to have an agent do things for you, again with no precedent.
— Ian SilberHiring designers doesn't require an AI background, it requires curiosity and prototyping
Ian says he's explicit with candidates: he doesn't expect an AI background, because the field moves too fast. What he actually looks at is three things — deep curiosity about AI, willingness to jump in and learn, and prototyping ability. Prototyping was always something he valued, but the bar is lower now, so it's easier to stand out. Third is strategic thinking: having your own point of view on where the tools are heading. He also stresses that the team has to be ‘round’: you need generalists, but also extremely strong visual designers, a brand team, prototyping experts, product thinkers with a strategic bent, because no single designer can satisfy all of it at once.
— Ian SilberSystems thinking matters because you're building composable blocks
Ian says OpenAI is designing superintelligence, and there are so many capabilities that you have to think clearly about the underlying primitives so they can stack on each other. He calls out what Notion did well early: build the blocks first, make everything composable. That way the product feels like one coherent, simple thing to users, and the model can reason about how to assemble those capabilities. His concrete advice to the team is ‘do less’ — if you can avoid designing something, avoid it; first check whether an existing system or component can be reused, or whether the feature is needed at all, or whether extending an existing feature is enough. That requires stepping outside the small piece in front of you and looking at the whole system.
— Ian SilberThe traditional design process has no time left, but some things need grinding
Jenny Nguyen has said the traditional process is prototype, mock, test, user research, iterate — and there's no time to run it all now. Ian says that is indeed how they work, but it depends on the stage: for some ChatGPT features they'll try a hundred, throw away ninety-nine, and ship one, and the composer changes every day; for others, like Codex or the super app, it's try fast, build in public, get internal and external feedback immediately. His test is ‘the technology might change in a month’ — some interaction or setting may not need to exist anymore, or latency changed, or intelligence changed, so you can rethink it. So don't sweat the things that haven't hardened yet; wait until they harden, then make them as good as possible.
— Ian SilberThe blank box problem is really about an interface that can morph
ChatGPT's user spectrum is absurdly wide: on one end people asking what to eat tonight or what haircut to get, on the other people in Japan automating a farm with AI. Ian says this creates a core design problem — the blank box problem. Critics say the chat box is just a terminal, a prettified command line, but the question is: how do you design an interface that can morph into anything? Some people come to generate images for fun, some to generate a deck. At the same time you have to serve two kinds of people: newcomers need to understand how this thing brings value to their life or work, while power users want you to get out of the way. Kevin Weil's point is cited to explain why the chat box hasn't been replaced — humans talk to both low-intelligence and high-intelligence beings, and speech itself is the best interface across the intelligence spectrum.
— Ian SilberThe super app's endgame is one unified entry point that requires no choosing
Ian describes the direction: ChatGPT will increasingly adapt to the full context of your life, and proactivity is the biggest unexplored part — it can already give decent suggestions from external inputs like your calendar and Slack. Voice just went through a model-level leap and feels much more natural. The super app's endgame is ‘one universal input’: you ask it, and it decides on its own whether to answer instantly, chat, or go off and do the work for you. Users shouldn't need to know which model they're using or at what speed. He also mentions wanting ChatGPT to support reusable, persistent workflows, because a lot of usage right now feels like starting from zero every time.
— Ian SilberGroupon taught him brand personality and company DNA can't be changed
Ian says early Groupon was a brand-driven product, and Andrew Mason injected a strong personality into it: the company had a comedy team, and the writing team was basically all comedians, and the product didn't take itself seriously. People loved reading that email not because of the discount but because it was funny and had personality. The bigger lesson: every company's DNA is different, and the way you work has to adjust to it. Groupon was one playbook; Instagram was design- and creativity-driven, with Kevin Systrom himself a designer; OpenAI is a research lab, and Sam Altman operates in a completely different way. He stresses in particular: you can't change the culture, especially at a company with a strong founder imprint — what you can do is understand it, work with it, and round it out.
— Ian SilberIn their own words · checked verbatim
someone who might be, uh, sort of classically trained in a certain way of working now feeling like, oh my gosh, I need to like really change everything.
Ian Silber4:07
And so I think what's encouraging about this is that nobody should feel like behind right now. I think that if you literally started today, you're going to have a leg up on pretty much most people.
Ian Silber12:25
Maybe one way to put it is I think it already is an incredible product designer. It really is actually already great.
Ian Silber22:52
It's like all of a sudden, you know, take the iPhone. We're designing for multi-touch. And that opened up like this whole new interaction paradigm. And there's no training data for that, right?
Ian Silber25:59
where I'm like, just do less. Like my encouragement has been do less. They're like, don't design if you don't have to, maybe there's already an existing system or component or something that we can build on top of.
Ian Silber32:15
The challenge, of course, is, like, not necessarily over-indexing and sort of, like, putting something that people might not understand in front of a billion users.
Ian Silber40:41
what is an interface that is very good at spanning a wide spectrum of intelligence? It's actually just talking.
Lenny43:52
you're not going to change that. You're not going to like come in and change that culture because that's very hard, especially if it's very founder oriented.
Ian Silber1:08:28
Figures
| ChatGPT monthly active users | billion-scale | 38:30 |
| Ian's years at Instagram | 8 years | 38:30 |
| Ian's years at OpenAI | 3 years | 0:00 |
| Try-to-ship ratio for polishing ChatGPT features | try 100, throw away 99, ship 1 | 33:17 |
| Idea-to-launch time for fast-iteration projects | 4 hours | 36:23 |
| Envisioned designer/engineer ratio at a startup | two designers to one engineer | 15:37 |
Glossary
- blank box problem
- The chat box has no fixed form; how do you design an interface that can morph into any tool?
- capability overhang
- Model capability far exceeds the part actually exposed to users in the product; most people only use a small slice.
- building in public
- Don't wait for a finished product — ship as you build, get internal and external feedback fast, then iterate.
- systems thinking
- Don't design isolated features; build underlying blocks that can stack on each other.
How to listen
Product designers and design leads anxious about being replaced by AI, plus PMs who want to know how OpenAI weighs a billion users against frontier features.
The opening two sections of mood-survey setup and the two mid-episode sponsor reads can be skipped.