OpenAI’s Head of Design: This is the best time in history to be a designer | Ian Silber

My tech-workforce survey found designers and researchers are the unhappiest group: more anxious, overwhelmed, and tired, and less likely to recommend the career. Yet another half are having the time of their lives. I wanted to understand that split with Ian Silber, OpenAI’s head of product design, formerly at Instagram and Artifact.

That uncertainty is totally rational. Designers trained in a traditional craft wonder, “Do I need to code every day now? Do I need to change everything?” AI speeds up ideas, but it doesn’t remove feedback, alignment, or judgment. You still try something, miss, and learn from users or teammates.

At OpenAI, we don’t want one prescribed process. We hire for curiosity and leave room to make bad ideas cheaply, then discard them. The designers embracing these tools feel more expressive and agentic: prototype immediately, use an agent at each step, and revisit tools that may work now. We’re incredibly early.

I think this is the best era ever to be a designer. An idea can become a working thing almost immediately. As software gets easier to make, the human side matters more: understanding people, making something delightful, and having a point of view. Boundaries blur, but larger teams still need product direction, design craft, and engineering rigor.

AI is already a very good product designer, though it won’t always be the best typographer, interaction designer, or visual designer. Humans still matter in watching people, spotting unmet needs, and inventing interactions without precedent. Multi-touch, Instagram, Snapchat, voice, and agents all began with asking, “What could this capability mean for people?”

I hire for a rounded team, not one mythical person who does every job: visual and brand craft, prototyping, product thinking, and frontier obsession. AI experience isn’t required; curiosity is. Systems thinking matters too. Use composable building blocks, not disconnected one-offs. Sometimes the advice is: do less—reuse a component, extend what exists, maybe skip the feature.

The process depends on the bet. For durable, personal parts of ChatGPT, we test, research, prototype, dogfood, and sweat countless versions. Elsewhere, we take a big swing, release quickly, and learn in public. Don’t get precious about something technology could invalidate next month. The interface must serve novices and power users without overwhelming either.

Chat is powerful because talking is natural across a huge range of intelligence, but it won’t remain text in a box. We’re exploring ways to state intent without prompt expertise, offer tappable follow-ups, and directly manipulate writing or images. It should adapt to context, become more proactive, use voice naturally, support richer outputs and durable workflows, and hide complexity when it isn’t needed.

If you’re stressed about design, that feeling makes sense. Then play. Prototype more, try a new tool, and focus on the outcome rather than worshipping process or token count. You still steer the work. I riff late at night, bring rough prototypes to the team, and use these tools for message summaries and meeting context.

I’ve had plenty of flops. With IGTV, we held onto bad assumptions, learned, changed course, and Reels became successful. What matters isn’t whether one launch misses; it’s whether you listen, adjust, and keep moving. At OpenAI, I felt intimidated and like I was failing daily. Peers reminded me those problems weren’t uniquely mine.

My vibe is: have fun, try stuff, stay flexible, build great things. Groupon showed me that character can make a product memorable; Instagram’s deliberate simplicity fit its DNA, while OpenAI calls for more experimentation. Learn a company’s strengths and adapt. Please use ChatGPT and Codex, send feedback, and show how your process is changing. That’s how we all learn.