Research run by agents
One agent analyzed a file of real user queries, formed hypotheses and wrote the report. A second pass tested new hypotheses from interviews against the data. A designer judged every step.
We are redesigning how people manage dozens of agent workflows in an AI browser, using research run largely by agents.

Polar is an AI browser. You type what you want and it does knowledge work across any website using your own accounts. A single request can run for hours. Most users are non-technical people in sales, recruiting and operations.
Workflows are Polar's most powerful feature, and as the co-founder put it, people just don't know that it exists. Power users run 20 to 30 workflows and can't tell what's running, what needs them, or what to delete.
Progress should be understandable, not just animated.
Research run by agents
One agent analyzed a file of real user queries, formed hypotheses and wrote the report. A second pass tested new hypotheses from interviews against the data. A designer judged every step.
A home control center
What needs your input, latest results, and suggestions in one place.
Projects that separate the work
Recurring workflows and what they did today, kept apart from one-off tasks.
Search and foundations
Full-page search with previews, the active account shown wherever a task runs, and a design system built from the codebase.

Before kickoff we audited the product and came with quick wins. The research reports went into Claude Design to generate new directions without opening Figma. Our designer built the sidebar options as a live code sandbox.
Next · VondyA creative platform grows into a building platform.Building something like this?
Start a project