Speaking Bucharest
Stop shipping what nobody asked for
I closed Day 1 of How to Web on the main stage with a live demo of rapid product discovery. The room voted twice, I built the prototype in Lovable, and Uxia's synthetic users tested it.
- Event
- How to Web
- Where
- Face Convention Center, Bucharest
- When
- 7 October 2026
- Format
- 90-minute live demo
- Recording
- Not published yet
The conference
How to Web, Bucharest
How to Web is Eastern Europe's leading startup and technology conference. The 2026 edition ran from 6 to 8 October, under the theme "What will you build when you can build anything?"
3,000+
attendees
40+
countries
250+
investors
Main stage
Build & Grow, last session of Day 1
Figures from How to Web.
The session
Photos: Mihnea Ratte.
What I did in the 90-minute live demo
- 1Context
Set the context: everyone is building now, and it has never been faster to ship confusion. The real question is how to decide what's worth building.
- 2Vote 1
Ran a live vote on AhaSlides: which challenge do I build against?
Help Boris, who tracks his calories, record his homemade dinner in 10 seconds Won81%Help Anna practice speaking Polish so she can order dinner in Warsaw next week 19%52 votes
- 3Vote 2
Ran a second live vote with the same tool: how will we try to kill the idea? The winner decided what I built.
Retention test: tests the need Won74%Mock sale: tests the promise 26%43 votes
- 4Live build
Built the first version live in Lovable and showed it on the big screen: Plateful, a calorie tracker. Snap your meal, AI estimates calories and macros, you approve. Some people in the room built along.
Open the app → - 5AI test
Tested it live on stage with Uxia's synthetic users: faster than testing with people, and it doesn't replace them. The app scored 77 on the System Usability Scale, where the industry average is 68.
Read the test report → - 6Iteration
Read the test results and fixed the prototype on the spot, so the room saw a full iteration live.
- 7Summary
Closed with a summary of the whole method and what we learned, including three Mom Test rules for talking to customers.
What the room took away
People left with learnings on how to run rapid product discovery.
Every idea is an assumption. Pick the go/no-go test before you build.
Behavior beats opinions. A tap on Pay is a commitment you can count. A compliment costs nothing and carries no data.
Build a good-enough first version fast, then iterate with synthetic users. They find usability issues, but they can't prove real demand.
Customers decide what's worth building. Ask about their life and the past: "When did this last happen?" beats "Would you use an app that…?"
Thank you to Monica Zara and Ada Berea for the invitation, and to the How to Web team for looking after the speakers so well. And thank you, Uxia, for partnering with me on this session.
Happy to run this live demo at your event.