Maximillian Piras is Yutori’s founding designer. His product design and writing examine how people can understand, direct and evaluate AI systems, including agents that navigate websites and complete tasks.[1][2]
Start with Using Friction As A Feature In Machine Learning Algorithms →
X · MVXMXM 
GitHub · MVXMXM Career
Piras studied graphic design at California State University, Long Beach from 2007 to 2011. At 8tracks, he moved from marketing work into leading product design. He later led design at Headliner from 2018 to 2024, helping podcasters share audio through automatically generated visual assets. His portfolio dates the Yutori founding-designer role to 2025.[1]
Contributions
In Mousepower, Piras asks how to measure whether agents produce useful outcomes when they can generate work faster than people can judge it. The talk distinguishes generation activity from value and uses uncertainty about a task’s steps and acceptance criteria to help decide what can sensibly be delegated. His earlier Bitter Layout talk examines why familiar prompt boxes and model pickers persist despite their usability costs.[2]
Ideas
Friction that improves recommendations
In his 2023 essay, Piras argues that a faster interface is not always a better one. Confirmations can prevent destructive mistakes, while deliberate questions can supply the preferences a recommendation system needs. The useful design question is whether the information or protection gained justifies the extra step. His examples connect everyday interface decisions with the quality of the signals a learning system receives.[3]
Interfaces beyond a conversation
His 2024 essay examines why a language model does not require every interaction to become a chat. Visual controls can make available choices easier to discover and compare, while conversation can be useful when a user needs flexibility. Piras treats choosing an interaction style as a design decision tied to the task and the capabilities of the system.[4]
Tools, writing & talks
Writing, research & talks
Start here
Using Friction As A Feature In Machine Learning Algorithmssmashingmagazine.comExplains when an extra interface step can prevent mistakes or improve personalization.
Start with the concrete interface examples in this essay. It introduces his design reasoning without requiring experience building an AI agent.
Author[3]A talk on judging agent value and choosing tasks whose outcomes can be evaluated.
Speaker[2]Original source- Designing For AI Beyond Conversational Interfacessmashingmagazine.com
Compares conversational and graphical approaches to interacting with AI.
Author[4]
