Context
Ian Silber has spent three years as OpenAI's head of product design, after eight years at Instagram and a stint at Artifact, the news app from Instagram's founders. Lenny's own workforce sentiment survey found designers and user researchers report the worst scores of any role: most overwhelmed, least optimistic, and least likely to recommend their profession to newcomers. Silber offers a structural explanation for why design has lagged engineering's AI gains, plus his case for why this is still the best moment in history to become a designer.
The Big Idea
AI hasn't sped up design the way it sped up engineering, because the design process still runs on a loop of trying ideas, throwing them out, and gathering real feedback, and AI hasn't compressed that loop the way it compressed a coding agent's binary pass-or-fail check.
Silber's own team has watched engineers get 10x, sometimes 100x, more productive with AI while design has barely moved, yet his answer to the anxiety this creates is that the field is early enough that starting today still gives anyone a genuine head start.
Key Insights
Engineers got 10-100x, design didn't
Silber's internal team surveys found engineers seeing 10x, sometimes 100x, productivity gains from AI, while design's gains have been far smaller. The reason is structural: a coding agent gets a largely binary task, does this code work, that AI can verify and iterate on by itself. Design has no equivalent check. You still have to try an idea, gather honest feedback, and often scrap it and start over. That loop of trying, testing, and discarding hasn't compressed the way code generation has, which is why the productivity gap between the two disciplines persists.
About half of designers are thriving
- Lenny's survey found roughly half of respondents having the best time of their careers, and the other half genuinely struggling.
- The strongest predictor is feeling amplified by AI rather than threatened by it, staying curious, and treating almost any idea as worth prototyping quickly.
- Silber's own habit: when an idea hits him at night, he throws it straight into an agent to prototype before taking it to the team.
AI is a great designer, not yet the best
Silber says AI already "is an incredible product designer," and uniquely accessible to anyone who wants to use it. But it isn't necessarily the best yet at visual design, information hierarchy, typography, or interaction design specifically. Humans stay essential at truly understanding what people need, watching how they actually use something, and inventing genuinely new interactions, the parts AI can't yet do as well.
New interaction paradigms have no training data
Multitouch on the iPhone and Snapchat's ephemeral messaging had no prior art for a model to learn from when they were first designed, because AI is trained on what already exists and a genuinely new interaction has no precedent in that training data by definition. OpenAI faces this now designing how people talk to AI by voice or hand real tasks to an agent, territory with just as little precedent, which is why human judgment matters most exactly where nothing has been tried before.
Systems thinking is now a core design skill
OpenAI is building many capabilities on top of each other, so designers need to think in composable primitives rather than one-off features, citing Notion's building-blocks approach as a model. Engineers are shipping fast enough that an ad hoc, feature-by-feature design process cannot keep pace. Lenny notes this is the fifth podcast in a row where systems thinking has come up as a skill people are actively seeking.
Mental Models & Frameworks
Capability overhang
The gap between what a product can do and what most users actually touch. OpenAI deliberately keeps the mainstream ChatGPT experience simple, while giving power users, through the desktop app and Codex, the cutting-edge capability first. Silber's example makes the range concrete: someone asking what a rash on their finger means sits on the same product as someone automating an entire farm in Japan.
Pick your durable battles
For the few surfaces confident to matter for months, like the chat composer, run the full obsess-test-iterate cycle: try 100 things, ship one. For everything else, build in public and take fast public feedback instead of the traditional research-prototype-test-iterate process. Lenny's summary, which Silber agreed with: you pick your battles and focus rigor on what's durable, even though it's genuinely hard to know in advance what that will turn out to be.
Trade-offs & Nuance
Roles blur, but responsibilities don't merge
At a startup, Silber would hire the best generalists who can move fluidly between design, product, and engineering. At a bigger company, he still wants someone accountable for direction (PM), experience (design), and system integrity (engineering), even as the underlying skills overlap more. His framing: the responsibilities stay real even as which person holds which skill becomes more fluid, so roles blur without actually disappearing.
Practical Application
Prototype ideas in an agent first
Silber's own workflow: an idea at night goes straight into an agent for a rough prototype before he takes it to the team. Treat this as a default first step for testing any idea, not an occasional shortcut reserved for when you're stuck.
Hire specialists, not one generalist designer
Silber hasn't found one designer equally strong at visual craft, prototyping, strategic thinking, and brand. Build a well-rounded team of specialists deliberately, instead of holding out for a single designer who covers everything, since that person basically doesn't exist.
Separate durable surfaces from exploratory ones
Decide explicitly which features are stable enough to deserve the full research-prototype-test-iterate process, and which belong to the larger, faster-moving surface that should ship and get thrown away without regret. Treating everything with the same rigor slows down exactly the work that's supposed to move quickly.
Learn a company's culture before importing one
Groupon's brand-led humor, Instagram's design-first simplicity, and OpenAI's research-lab culture are genuinely different operating styles, not variations on one playbook. Especially at founder-led companies, work to understand and round out the existing culture rather than trying to import what worked somewhere else, since that transplant rarely takes.
Questions to Consider
- Where does our own product's process lack a clear yes-or-no check, the way code either works or doesn't, that would let AI iterate on it directly the way it does for engineering?
- Are we exposing our product's full advanced capability to everyone, or deliberately keeping a simple default while making that depth available to the people who go looking for it, the way OpenAI does with ChatGPT versus its desktop app and Codex?
- Which of our features are genuinely durable enough to deserve the full research-prototype-test-iterate process, and which are we over-investing that same rigor in even though they're likely to change soon?
- Do we actually understand our own company's culture, or are we quietly trying to import a process that worked somewhere else instead of building on what's already here?
Bottom Line
Nobody is actually ahead right now. The field is early enough that starting today still gives you a real head start, and the anxiety many designers feel is a rational response to genuine structural ambiguity, not a sign they're falling behind.
Focus on the outcome you're producing, not the tooling around it.
Case Studies Mentioned
Instagram's IGTV to Reels pivot
IGTV, Instagram's long-form vertical video product, was a huge flop built on assumptions the team held onto too long. The team changed course quickly and shipped Reels instead, which became a genuinely successful product. Silber's lesson: what mattered wasn't getting it right the first time, it was how fast and honestly the team reacted and kept iterating.
Groupon's brand-led daily deal
Groupon, Silber's first design job under CEO Andrew Mason, was built by a team that included a comedy writing staff. People opened Groupon's daily deal emails as much for the character and writing as for the discount itself. The lesson Silber carries forward: a strong, distinct brand identity can differentiate even a fairly ordinary commodity product.
People to Follow
Kevin Weil
Former Chief Product Officer at OpenAI. Argued that chat works as an interface across a huge range of user sophistication because talking is naturally something humans already do across a wide spectrum of intelligence. Also the source of the line "the model we have today is the worst the model will ever be."
Resources Mentioned
| Resource | Type | Why it was mentioned |
|---|---|---|
| The Design of Everyday Things, Don Norman | Book | Silber's go-to design recommendation, still a classic despite not being a unique pick |
| "Stop the AI Confidence Theater" | Post | Growth leader Elena Verna's critique of people performing AI mastery online; Silber agrees it's better to be humble |
Notable Quotes
"It already is an incredible product designer. It really is actually already great." (Ian Silber)
"The model we have today is the worst the model will ever be." (Kevin Weil, former OpenAI CPO)
"We're all extremely early in this process. Nobody should feel behind right now." (Ian Silber)