Fearn has raised $5.5 million from Kindred Ventures, a16z Speedrun, Designer Fund and Essence VC to build an AI-native patent prosecution firm for startups.
The pitch is simple: bill for outcomes, not hours.
Co-founders Han Kim and Angela Gao, both Caltech alumni, started by building software for law firms and then walked away from it. Kim had prosecuted patents at Morrison & Foerster and Gao holds a PhD in computer science and AI, and they concluded tooling alone cannot fix a model where time saved is revenue lost and pre-filing IP cannot safely touch generic models. FearnOS is their answer. It represents a patent as a graph that links claims to supporting text, figures and technical material, preserves attorney edits and keeps a full production record, then runs specialized models with deterministic checks before a former Big Law patent expert reviews every application. The firm says work that typically takes 30 to 40 attorney hours can take as little as 30 minutes on FearnOS, with provisionals in as little as three business days, a $2,500 provisional and $9,000 non-provisional fee including USPTO fees, gross margins above 80%, and a guarantee that puts the drafting fee at risk if a non-provisional receives no allowed claims.
Early users show why speed matters. London game studio Iconic moved from disclosures to filed applications in days for AI characters that improvise in real time. Defense tech company Photon Spear filed a hardware space technology patent in several days while keeping material entirely on privately hosted infrastructure. Serova Bio is using the platform for AI-designed personalized cancer vaccines where claims and disclosure must stay coordinated as the science evolves. "Speed is the whole game at seed stage," said Andrew Bowell, CEO of Iconic, and Daniel Mukasa, CEO at AbInitio Bio, said Fearn got him to a strong first draft in a weekend.
The context is not just another legal AI tool. The first contrast is the traditional Big Law prosecution model that Kim trained in at Morrison & Foerster, where a single application often costs $18,000 to $40,000 and takes months. The second is the software-only approach Fearn itself tried first, selling better drafting tools into firms that still monetized hours. Fearn is betting founders will choose a firm built around the technology instead, while major firms are separately spending heavily on proprietary AI infrastructure because sensitive legal work demands control over models and client data, a trend Fearn cites to argue startups should get that infrastructure without the Big Law economics.