some of the most interesting innovation has always been at the edges. I did my phd in a highly interdisciplinary field called computational imaging where I worked with generative models to solve real world scientific imaging problems. I worked with domain experts to blend messy data with mathematical rigor.
my biggest takeaway wasn’t really technical. it was that the most interesting problems live between fields and you can’t shortcut your way there. you need to get real depth in both domains.
near the end of my phd, I got really excited about program synthesis despite not having worked on it at all. I found it so deeply satisfying that there were fields of AI where you could guarantee whether or not you were correct (shout outs to @atharva_sehgal who introduced this to me) but most fields of science can’t be confined to strict logic.
that asymmetry is everywhere. AI is sprinting towards math because verification is built in. but in other fields, like in geophysics, biology, law, and medicine, it’s so much harder to tell if a hypothesis is actually correct. being able to quantify this is one of the biggest unlocks
it’s part of why @hanhanhan_kim and i started Fearn, building verifiable programs for IP law. we’re grounded upon the foundation that interesting work is at the edges where rigor hasn’t shown up yet
Patent systems are first-to-file. It's produced a quiet two-tier system: well-funded companies file in 48 hours. Everyone else waits weeks.
Today we're announcing @fearn_ai's $5.5M seed to end that gap. 🧵