A former Perplexity engineer behind the Comet search tool has raised $5.7 million to build a browser that completes hours-long knowledge work tasks with minimal human input. Polar, founded by Kevin Jiang, Vishaal Ram, and Howard Zhong, emerged from stealth this week with seed funding led by Madrona and angel investors including former GitHub CEO Thomas Dohmke.
Unlike traditional AI browsers focused on summarizing pages or answering questions, Polar interacts with websites the same way a person would. It clicks buttons, fills forms, scrolls through pages, and logs into accounts to complete multi-step workflows spanning research, recruiting, sales, operations, and marketing. The company reports that customers have already run over 4.5 million actions through the platform during its first seven months.
“Almost all knowledge work runs through the browser, and until now nothing could do that work for you,” said Jiang. The startup says its browser agent outperforms systems from OpenAI and Anthropic on browser-agent benchmarks, and has completed individual tasks lasting more than 15 hours without human intervention.
The funding arrives as tech giants and startups race to build AI agents capable of performing real work inside everyday software. Investors see the browser as one of the largest untapped opportunities for AI-driven workplace automation.
