A designer who doesn't ship code, on how a 30-day experiment with 500 people turned her into someone who ships products
500 people vibe-coded for 30 days. I was one of them. - Sanja Grbic, Automattic · Sanja Grbic
18 min total·Actually worth watching closely: ~5 min·3 must-watch clips
- 0:01 – 2:10Listen
A designer who had never shipped code
Sanya introduces herself and sets the premise: she's a product designer of over a decade who had reviewed PRs and played around with Claude Code, but had never actually built and shipped something in code at work. Automattic launched Radical Speed Month, a one-month initiative where teams paused their roadmap work.
The narrator's starting point is 'has never shipped code' — everything that follows should be read against that premise.
She's mostly on stage talking about her background and where the experiment came from; there's nothing to look at, so listening while you do something else is fine.▶ Jump to 0:01Speaker · Sanja Grbic - 2:10 – 5:00Listen
Pave the road first, then let people run
What Automattic laid down so everyone could actually use AI: encouragement to use, learn and experiment with it from the very beginning; an AI enablement project where every single employee goes through a two-week course designed specifically for their role, completely immersive with lectures and focused hands-on time; and a systems and operations team that set up the security, processes and documentation, on top of a couple of decades of very well documented work.
Company-wide speed isn't produced by slogans — it's paid for with training, security practices, and documentation.
All organization-level practice over plain bullet slides — following the reasoning matters more than watching the screen.▶ Jump to 2:10Speaker · Sanja Grbic - 5:00 – 7:03Watch
The rules, and the first two-hour shot
How Radical Speed Month worked: pause the roadmap work for 30 days, pair with a partner, and ship something real; AI was welcome but not a requirement. The first round had about a third of the company — around 500 people — starting around 794 projects. Then the first project: a board game session app four people built in two hours.
The toughest rule is 'ship something real' — it's what separates the experiment from playing around.
The second half cuts to the finished first project — create, view, join and manage board game sessions — and one look tells you what density of thing two hours actually produces.▶ Jump to 5:00Speaker · Sanja Grbic - 7:03 – 10:10Listen
The engineer becomes an enabler, Figma steps back
During the pairing the engineer didn't write her code — he set the project up in GitHub, helped everyone grasp the basic commands and understand how Git actually works, so designers who aren't comfortable in code could commit too. Meanwhile her design process inverted: the main artifact went from high-fidelity Figma work to the working prototype — a good plan first, then build directly, with Figma coming back later for a mood board and UI fine-tuning, and even handing Claude Code an image to convey visual intent worked.
The leverage an engineer gets from enabling others may be far greater than from building it themselves.
This is her looking back on the process change — carried by narration, with no demo you need to watch.▶ Jump to 7:03Speaker · Sanja Grbic - 10:10 – 11:40Watch
Doubted, then she built it anyway
She proposed a design system status tracker; the engineer raised questions about performance, about maintenance, and whether they'd be able to build it in the first place. She built the full idea as a proof of concept on her own and shows it running.
When the bar for delivering drops to what one person can clear, a feasibility debate is often worth less than just building the thing.
What's on screen is the finished product: links picked up from the GitHub repo, Storybook and Figma, sorted by status, tagged with the correct library names, search that works, live component previews — details that only carry if you watch.▶ Jump to 10:10Speaker · Sanja Grbic - 11:40 – 13:23Listen
From designer to design engineer
She defines what the shift really meant: AI enabled it, but most of all it came from owning the entire process and having the agency to make every decision, without the negotiations and handover that make this so hard in large organizations. And the bar for 'design engineer' is being able to push code to production in a well-established large system that demands serious onboarding and security practices.
What made her stronger wasn't the model — it was no longer having to hand her ideas off to someone else.
The densest stretch of argument in the talk, purely spoken — worth rewinding and hearing again rather than watching.▶ Jump to 11:40Speaker · Sanja Grbic - 13:23 – 15:00Watch
Two designers, six days, one complete app
The third project: in only six days two designers went from zero to an iOS chat for WooCommerce merchants that lets them answer shoppers in real time, reply from their phone, or set up AI to answer for them. She notes that alignment and ideation were easy for two designers, because they always have the users in mind.
Pairing doesn't have to be cross-functional — two people in the same role can align remarkably fast in the concept phase.
The demo unfolds every layer: authentication, the Jetpack connection, a widget that inherits the site theming, and an agent that scans the site, answers visitors in real time and discerns when it can't answer. Watching it is what makes the number 'six days' land.▶ Jump to 13:23Speaker · Sanja Grbic - 15:00 – 17:46Listen
Reproduce it on your own team
The closing, practical advice: changing processes or tools in a large organization means shifting the human behavior behind them, so provide access to the new tools, find your enablers and champions, create space for experimentation and give people agency. For practice, pick a project unrelated to any existing product so there are no risks involved and the focus is learning. And start the exploration as a Claude Code project folder, with all the chats and ideas recorded into a file system.
The sign that speed is really unlocked isn't one person getting faster — it's the team leveling up together with a shared understanding of how to apply the tools, so they can easily meet each other. That step doesn't need company approval; try it at your team level.
A spoken checklist of recommendations over bullet-text slides — listen and note the two or three you can actually use.▶ Jump to 15:00Speaker · Sanja Grbic