To err is AI — show & tell
AI makes mistakes. Some are frustrating. Some are hilarious. Some make us question reality. Let's bring them all to light.
We're past the era of six fingers and counting letters in "strawberry." As models get smarter, their mistakes get weirder — more nuanced, more human, and way more interesting. Especially when code, reasoning and tool use are involved.
This is a community show & tell. Bring your best (worst?) AI fails from your own experience. Are there sporadic issues, or is AI totally useless for your use case?
We want the real stuff: the bugs that made you laugh, the outputs that made you cry, and the moments that made you rethink how you prompt.
What to expect:
• A short talk by Piotr Migdał from Quesma on the evolution of LLMs.• Lightning talks featuring the most interesting AI mistakes from the community (send us!)• Discussion on where and why models fail — and how to work around it • Tips on prompting, double-checking, and knowing when not to trust the output
Want to present your AI fail?Fill out this form: https://forms.gle/7R3dBsSV3S9WNJY86 or simply apply directly to the organisers. Screenshots are strongly encouraged.
Entry details
The payment for ticket is not mandatory, but much appreciated! Helps us keep going and get you nice snacks!
Organised by Kolektyw3 and Quesma
About Quesma
At Quesma, we craft the hardest engineering challenges to test the world’s smartest AI.We provide leading AI labs with tasks like CompileBench and OTelBench. And we are now hiring - more details at the event!
About Kolektyw3
Kolektyw3 (K3) is Warsaw's Web3 and AI community hub — a coworking and event space at Koszykowa 54, built by builders, for builders. K3 combines everyday coworking with a regular lineup of meetups, workshops, demo nights, and hackathons, creating a space where founders, developers, and creators actually meet, collaborate on real projects, and support each other.
