Zebra
Talk to a zebra in the language you are learning, then get a summary of the phrases you got wrong and how to say them. Payment step connected.
~1hof building, to a working voice app per team
Case study · Warsaw · 27 October 2026
Poland Tech Week 2026Imaguru Startup HUB3 hours
3 teams.
3 apps you can talk to.
One challenge, one test, and the number that would kill the idea, all set before anyone built. Then about an hour of building with AI, a synthetic usability test and a three-minute pitch.
3 of 3teams pitched a live app with the paywall built in
8/100Upgrade clicks needed to keep the idea alive, set before building
3 minper pitch, with the live app and no slides
What they built
Three answers to one sentence.
The challenge
"How might we help Sofie, who has been learning Spanish for 600 days but still cannot say a sentence out loud, hold a two-minute conversation before travelling to Barcelona?"
Modeled on a paid feature that exists: Video Call in Duolingo Max ↗
🏆 Winner
What the evening showed
Building took an hour. Deciding is the hard part.
A price on the screen beats a debate.
The room asked one team why their plan cost so much, and they defended it with their own subscription bill. The better answer is a test: two prices, real users, count the clicks.
On a full dayYou leave with a costed plan to test it on real customers, and the number that means go.
Deciders buy products, not prototypes.
Anastasiya Khamiankova, co-founder of Imaguru, pushed every team to stop presenting a prototype and present a product. The winning team ran out of time before it reached its paywall.
On a full dayThe three-minute case is rehearsed with questions first, and the day ends in go, pivot or stop.
Synthetic users need lead time.
AI users take five to forty minutes per run. Most Warsaw reports landed after the pitches, too late to change what the teams showed.
On a full dayThe panel runs while you work, then you fix what it found and run it again.
Who ran it
Marcin Klimaszewski
I help product teams stop guessing: take a stuck idea and turn it into a decision you can trust, fast.
15 yrsin product and UX
100+workshops across Europe
150+people trained in 6 countries
35people in the team he hired as Head of UX, across five countries
Former Head of UX at Sanoma Learning, with a team of 35 hired across five countries. Certified Lovable Expert. Rated 4.7 out of 5 on Trustpilot ↗
Run with
The tools in the room
Back at Imaguru, with the same two tools
Lovable and Uxia were part of the first open Product Build Day here in July. For Poland Tech Week they came back and covered the tools for everyone who built.
Testing tool
Uxia, from Barcelona, tests apps with AI users. Each one thinks out loud as it clicks, so you can read why it got lost. By the company's own count, more than 900 product teams use it.
On the day
Uxia gave every team a full account for the evening, enough to run a full panel of AI users.
A challenge about a trip to Barcelona, tested by a team from Barcelona. Thank you, Victor and Borja.
uxia.app ↗Build tool
Lovable builds working web apps from a description in plain words.
On the day
Every seat got 100 Lovable credits, to keep building after the evening ends. The tip of the night: to move a feature between accounts, ask Lovable to describe it and paste that into the team's project.
Thank you, Lovable, for July and for October.
lovable.dev ↗The room and the test
Five AI users went through every app
Among the builders were founders and product managers, a designer, and people from HR and operations, working in three teams.
The build
Everyone built alone in Lovable first, then three teams merged the best parts. About an hour later, all three pitched a live app you can talk to, with the paywall built in.
The brief to Uxia
Each team published its app and gave Uxia a short task, such as finding the speaking feature and upgrading. The challenge went in as the scenario, and the last screen of the flow as the place to stop. Uxia can build the audience straight from a website link.
The report
Each AI user thinks out loud as it clicks. A run takes five to forty minutes, so most reports landed after the pitches, ready for each team's next version.
Roles from the applications.
Run it with your own team.
Teams · in-houseOne day in your office, up to 20 people, against your own challenge. It ends in go, pivot or stop.See the Teams day →Pro · Berlin, 26 NovemberOpen one-day training in a curated group of 12. Send one product manager or designer, and they come back with the method.See the Berlin training →
Or book a free 15-minute call ↗