OpenAI’s Head of Design: This is the best time in history to be a designer | Ian Silber
In a Nutshell
OpenAI's Head of Design Ian Silber says designers are the most anxious, overwhelmed group in tech because the industry is shifting faster than they can adapt. While engineers gained 10-100x productivity from AI, design still requires messy iteration, user feedback, and alignment—but designers who embrace AI tools report feeling amplified rather than replaced. The core message: this is the best time in history to be a designer because human creativity, strategic thinking, and understanding unmet needs are becoming more valuable as AI makes product creation accessible to everyone.
These notes were generated by AI and may contain inaccuracies.
Designers and user researchers rank as the most unhappy group on every dimension in sentiment surveys of the tech workforce. They report feeling the most overwhelmed, anxious, least optimistic, most worried, and most tired. They are also the least likely to recommend their role to people entering the workforce.
The lack of clarity around what is expected of a designer creates significant unease. Designers who were classically trained in certain ways of working now feel pressure to completely change everything about their approach. This uncertainty stems partly from observing that engineers have already gone through a major shift in how they work, while the design process remains unclear.
Engineers have achieved 10x or sometimes 100x productivity gains through AI tools, but design teams have not experienced equivalent compression. The design process still requires trying multiple approaches, throwing out ideas, and going through messy feedback loops. While AI can compress the time to generate ideas, the fundamental need for iteration, user feedback, and internal alignment remains.
The traditional design process of prototype, mock, test, research, and iterate no longer fits the current pace. At OpenAI, certain features receive obsessive attention with hundreds of iterations, while other features embrace building in public with big swings and rapid learning.
Ian Silber serves as Head of Product Design at OpenAI, a role he has held for three years. Prior to OpenAI, he worked at Artifact, the news app from Instagram's founders. Before that, he spent eight years at Instagram.
While the survey shows widespread unhappiness, many designers at OpenAI express optimism and excitement. The company culture treats the organization as a research lab, which encourages exploration rather than mandating specific changes to workflows. This approach reduces pressure on designers and allows them to maintain their existing processes while exploring new tools.
The design process remains inherently messy and fluid compared to the more binary outcomes from coding agents. Designers must still try ideas that fail, incorporate feedback, and iterate through multiple rounds. However, AI tools allow designers to generate and test more ideas faster while preserving the essential feedback loop.
OpenAI's experimental approach and research-driven DNA creates an environment where designers feel supported in exploration. The company seeks designers who may not have AI backgrounds but demonstrate genuine curiosity about tools and their applications. This supportive environment leads to greater creative expression and the ability to try bad ideas without pressure.
Designers who embrace AI tools report having significantly more ability to be creative, express themselves, and move through multiple ideas quickly. The bar for design quality is rising as thinking becomes stronger through tool usage. Designers gain greater agency and ability to have high impact.
Designers having the best experience report feeling amplified by AI rather than replaced. Key attributes include being very curious, having time to explore, and feeling more creative. They use AI at every step of the process to help generate ideas, prototype, and think through problems. Adaptability proves crucial as the industry shifts rapidly.
The design industry has shifted dramatically within six months. Tools that had significant limitations one month become highly functional the next. This rapid pace of change means that what didn't work yesterday might work today, fundamentally altering the design process.
The current state represents early innings of the entire AI evolution. Nobody has figured out the perfect design process, and whatever works today will likely change again. This early stage means nobody should feel behind - starting today provides a leg up on most people.
When Silber entered design, learning required viewing source code on HTML pages and using tools like Dreamweaver. Today's reality allows moving from idea to working version in minutes, making design more accessible and expressive than ever before.
As AI enables anyone to make products, the human element of design becomes increasingly important. Companies with great design stand out more. Products that truly understand users and create delightful experiences gain competitive advantage. Designers' core job of understanding users and creating delightful experiences becomes more valuable.
Traditional startup ratios of one designer to fifteen engineers may shift. Some founders discuss hiring two designers for every engineer, prioritizing creative thinking and product vision alongside solid engineering.
Designers at OpenAI are empowered to take early research concepts and figure out how to package them into products people want to use. This represents a designer's fundamental job of solving unmet customer needs.
Product managers, engineers, and designers all believe their role will dominate while others diminish. However, the roles are blurring together as they have always had overlap. Designers who excel at product management or engineering work exist, and single siloed role definitions don't fully capture reality.
Despite increasing overlap, distinct responsibilities remain valuable, especially at larger companies. Having someone rally multiple teams, sharpen problem focus, and handle resourcing provides value. One person cannot effectively handle all aspects of design, product management, and engineering simultaneously.
Startups benefit from hiring generalists who can fluidly move between roles. Larger companies benefit from specialists who can provide focused expertise while still having the ability to contribute across boundaries.
AI already functions as an incredible product designer that is accessible to every person. While it may not be the best at visual design, information hierarchy, typography, or interaction design, it continues improving rapidly. The focus should be on building tools that enable human creativity rather than replacement.
Human brains remain most valuable in truly understanding what people need through direct observation and feedback loops. The ability to invent something new and provide a unique human perspective becomes increasingly important. Having a clear point of view about who you're building for and why distinguishes human designers as AI capabilities advance.
The PM of Codex explained that AI is not good at design for several reasons. Great product design involves doing something new, but AI is trained on existing websites, so it reproduces what already exists rather than creating something novel. Additionally, AI researchers are not innately great at design and may not understand exactly what it takes to train a model to excel at this skill.
Great products like the iPhone introduced multi-touch interaction, which opened entirely new paradigms with no existing training data. Instagram embraced the reality that everyone had mobile phones. Snapchat flipped communication norms with its interaction design. These examples demonstrate that great design means truly understanding unmet needs and building solutions in unique ways. AI cannot replicate this because it lacks precedent for designing new interaction paradigms.
Humans excel at evaluating what good design looks like and iterating based on feedback. Having humans in the loop remains crucial, especially as AI capabilities evolve rapidly.
A well-rounded design team requires different strengths rather than expecting one designer to check every box. Teams need generalists who can do everything alongside specialists: incredible visual designers, brand teams, prototypers, strategic product thinkers, and those obsessed with cutting-edge developments. This diversity becomes valuable at both startup and larger scales.
Curiosity about AI matters more than prior AI experience, since the field evolves too quickly for anyone to have comprehensive background knowledge. Designers must demonstrate high aptitude for learning how to use AI in their process and how to create tools that enable more people to use it.
Prototyping skills have become increasingly important and accessible. Candidates who prototype effectively stand out during interviews.
Strategic thinking and having a point of view on where technology is heading are valued.
Systems thinking is becoming more critical. Building composable primitives that stack on each other, rather than one-off siloed experiences, enables cohesive products where models can reason across capabilities. Notion demonstrated this well with their building blocks approach.
Engineers ship thousands of PRs daily, reducing time for traditional design processes. The response involves thinking outside individual project bounds and across the entire system. The principle is to do less: avoid designing what already exists in the system, and consider whether features are even necessary.
Design approach varies by feature and product stage. For ChatGPT features, teams may try a hundred ideas and ship one after extensive iteration. Other areas like Codex involve rapid experimentation. Building in public allows instant internal and external feedback. The key is being comfortable with quick iteration rather than being precious about work until technology stabilizes.
The strategy involves picking battles and focusing effort on durable elements unlikely to change. For ChatGPT becoming deeply embedded in personal life, teams spend significant time thinking through the experience rather than shipping overnight. Other work moves from idea to ship in four hours. The challenge lies in predicting which elements will remain stable.
ChatGPT serves vastly different audiences, from users seeking dinner recipes to those automating farming operations in Japan. The concept of capability overhang means most users access only a sliver of true value, which is acceptable since they still receive substantial benefit. The approach involves creating extremely simple, streamlined experiences for the majority while making cutting-edge features accessible to those who care. Experimentation happens early with engaged users through desktop apps and Codex before features eventually reach the main experience.
A major challenge is the blank box problem: designing an interface that can shapeshift into anything. Users may come for image generation, PowerPoint creation, or other tasks. The interface must adapt to different contexts and provide appropriate affordances based on user type, whether novice or power user, data scientist or personal life manager.
Talking works as an interface because humans naturally communicate across intelligence levels. The design focus involves helping users express intent better without requiring perfect prompt engineering, teaching them what the model can help with, and understanding what they want through follow-up questions. Writing blocks allow direct manipulation of generated content like emails, moving beyond pure text transcripts.
The product will become more adaptive to life context. Proactivity represents an untapped opportunity, already visible in ChatGPT Work's calendar and Slack-based suggestions. Voice interaction will continue improving. The goal is a universal input that knows when to answer quickly, be chatty, or execute work independently. Users should not need to think about models, speeds, or technical details. Richer visual outputs that are more interactive will emerge, along with durable workflows and attributes that can be reused rather than starting from scratch each time.
The most important realization is recognizing how early the field remains. The role involves significant complexity in empathizing with diverse user types across free daily users to those paying hundreds of dollars monthly.
When GPT-4 came out, I was using it but didn't feel like I had a true understanding of it. There's still so much to learn. The idea that we're early and figuring this out is something people want to hear.
Stepping into a role managing at a scale that's growing fast at a company that's different means every day brings something new. For a long time, I felt like I was bad at this or didn't know what I was doing. Talking to friends who had led really large design teams or were in a similar transition revealed that the problems weren't unique.
At any stage of your career, there's that feeling of imposter syndrome - what am I doing, I don't know how to do this. Finding people you can talk to who are going through similar things helps because you'll find there's usually a lot of shared challenges and struggles. Having a mentor or an impartial buddy outside the situation was really helpful. Finding that community has been valuable.
Understanding that this is all so new and evolving really quickly is important. Elena Vera had a great post called "stop the AI confidence theater" where she pointed out that people keep posting all these things they've done with AI and how it's changed their life, but very few of these things are actually that great and working. It's better to be humble about it and embrace that it's changing quickly.
This connects to what Kevin Wheel, former CPO at OpenAI, said: the model we have today is the worst the model will ever be. A new model comes out every week or every day if you count all the labs. It's changing so quickly and everyone gets better. It must be like building or designing in the early days of computers when componentry was evolving quickly with drastic step changes - you'd hit limitations and then suddenly those limitations were gone.
For designers feeling overwhelmed, anxious, struggling, stressed, and uncertain about their future, recognize that it's a totally rational feeling. It's a great time to explore - what if I tried working differently, rethinking tooling, embracing prototyping more. Recognizing that we're all still so early and it's still changing, and embracing that. Have fun with it and see where it goes.
Focus on the output and outcome. At the end of the day, we're not here to build tools, processes, or token max - we're here to make great software that people love or great products. If you focus on that, these tools can help you do that, but you still have a huge role as the person steering and directing it.
One way AI has been useful is spending less time in the pixels and using it to help think through broader ideas. Having an idea at night and quickly typing it in to prototype something gives an idea to talk to the team about and helps visualize something. ChatGPT, whether it's ChatGPT or Codex, can help visualize early ideas or when riffing through something.
On the operational side, using it for checking and summarizing important Slack messages, giving context before important meetings, recruiting - every part of the process gets augmented.
ChatGPT has been incredible because it runs fully in the cloud, so you can be anywhere and talk to it and have it spend time and come back with a really great artifact. Being able to quickly visualize ideas before taking them further and play with them has been fun. The accessibility - not even needing to be near a computer - provides flexibility.
There are so many times of failure. Even at OpenAI, I've had to learn every step of the way. When I first joined, I didn't fully understand how I wanted to build a design team or how to hire. I felt like I was failing every day but quickly trying to adjust and learn. My whole career has been learning on the go, never planning any one step, quickly learning from mistakes.
I've shipped plenty of things that didn't work out. At Instagram, we had flops and quickly moved on. I worked on IGTV which was a huge flop, then we shipped Reels. We baked in some wrong assumptions and held onto a few constraints, but learned from that quickly, changed it, and then it became a really successful product.
This has helped harden me to critique and feedback. Sometimes you get it wrong, sometimes you try something and it's about putting it out in the world, getting feedback, and iterating. More importantly, it's not about what you shipped one time but how you react, fix forward, continue to iterate and listen to users.
People's careers are mostly failure along the way. It's helpful for people to hear that.
This is a really fun time to build stuff. Designers can be really adaptive and focus on making really fun, making really great stuff. Great companies are seeking designers and seeking great design. There will be companies left behind because they haven't shifted, but there's plenty of room for amazing, great design, and that's going to be more true than ever.
The Design of Everyday Things is a classic for thinking about how we think about design. It's the one recommendation that always comes up when asked about design books.
Maybe Happy Ending is a Broadway show in New York about super intelligent humanoid robots with incredible production, set design, and everything unique about it. The Invite is a play with four people in one beautiful apartment with amazing set design, writing, and acting - it's a wild ride that you go into knowing nothing about.
The Rivian EV is impressive because they've removed assumptions about how operating a car should work - you just walk up to it and it unlocks, you don't have to turn it on and off. Waymo is the most magical experience if it could be broadly accessible.
The approach is to have fun with it and just try stuff. Don't feel like you have to fit into a specific box. It's perfectly alright to be very fluid about the way you work or think about things - be flexible and adaptable. The vibe is just trying to have fun with it and build great stuff.
Groupon was a brand-led product under Andrew Mason with a comedy team and writer team. The product didn't take itself seriously in its early years. People loved the email because it was well written, funny, and had character. The company did a really good job with brand and embracing that side of things.
Every company has unique DNA and the way you work changes based on that. Groupon was run one way, Instagram was very design and creativity focused under Kevin Systrom, OpenAI is a research lab under Sam Altman with a completely different way of operating. Learning that companies have different DNAs and understanding how to embrace and leverage that has been valuable.
You can't come in and change culture, especially if it's founder-oriented. At Instagram, the motto was "do the simple thing first" and that's been carried through, but sometimes it's not the right way to think about things at OpenAI where we're still experimenting. Being flexible and helping round out the culture rather than trying to change it has been the approach.
Feedback and thoughts on ChatGPT and Codex are welcome. Available on Twitter for conversations with designers at all stages, engineers, PMs. Love when people show off how they're working and how that's shifting - showcasing process helps the community learn, especially as it becomes more opaque.
Personal website at iansilber.com with a deep dive into past work and an Easter egg game to discover.
Keep Lenny's Podcast in your library
Save the videos and channels worth coming back to, and find them again in one place.
