Gavin Nelson

The largest change is that I now work mostly in person. I moved back to the Bay Area recently and was excited to have that opportunity. I still value the flexibility and focus of remote work, but I also get energy from being around my teammates and find that many of the most useful conversations happen casually, outside of scheduled meetings. Sitting with other designers also lets me see what they’re working on and how they’re thinking about other parts of ChatGPT. That context helps me design my part as one piece of a coherent system.

Mostly I’ve cut everything back to the essentials and done some obsessive work to hide most cables from view. I find that keeping my space minimal and tidy helps me focus on the important things. A lot of the time I’ll have coffee and a notebook sitting on my desk while working, but yes, this is really what it looks like day to day.

I don’t really follow any defined process with regard to tools. It mostly comes down to thinking about what job I need to get done, and then picking the tool that is right for that particular job. Generally, for me, that means:

• Discussion, paper sketches, and whiteboard sketches for initially thinking through what it is I’m designing, what problems it should solve, the current constraints, etc.
• Figma for quickly drawing out ideas in high fidelity. I create a mess of divergent ideas, quickly iterate through layout, hierarchy, and visual design, and use Figma Slides to help share and describe my thinking.
• Codex for bringing ideas to life beyond a static image and creating various levels of fidelity of “the real deal”: gestures, behaviors, animations, transitions, haptics, and all the details that give something its final 20%. I’ll often pass Codex links to Figma to quickly get the base layouts right and go from there.

The order changes all the time. Sometimes I’ll bring a prototype I built in a few prompts with Codex to an initial discussion so we have a point of reference. Sometimes I’ll take a photo of a sketch, pass it to Codex, explore the pros and cons of different interaction models, and then turn to Figma.

I think AI has primarily broadened the term “design tool.” It’s now easier than ever to spin up a small piece of software for a very specific task, so you’re less limited by the functionality existing tools offer. The tools we have access to can greatly influence what we try to create, and AI makes a much wider range of possibilities accessible. Sometimes the right design tool is Figma, sometimes it’s Blender, sometimes it’s code, and now, sometimes it’s a brand-new, single-purpose thing you quickly build just for that one job. That flexibility is incredibly valuable.

I think a lot of the creative thinking and responsibility for taste will lie with designers for a while longer, even as the tools get faster and smarter. When you can design and build almost anything very quickly, a bigger part of the job becomes deciding what’s actually worth making: which problems matter, which direction feels right, what should be removed, and how something fits into the product as a whole.