FLF’s Epistemic Case Study Competition: Results

This summer, FLF ran a contest which asked entrants to help us “find the best workflows and methodologies for using AI to produce reliable, trustworthy knowledge bases, grounded in real-world cases”. We awarded just over $200k in prizes[1] and we are continuing to invest in this area (you can subscribe for updates). Why we ran this contest As we wrote in the announcement, > The heights of human epistemic investigation are impressive and valuable, but rare and difficult to reach… The limiting factor is rarely exquisite insight (though this helps!), and more often diligence, a curious and open mindset, and the time and effort needed to do the thorough work investigating background on a topic: activities AI is well placed to assist with. It’s still, in 2026, disappointingly difficult to track the sources and justifications for what you’re seeing, in order to check those — or similarly to see who else has checked, and corroborated (or argued against). It needn’t be so difficult! AI chatbots and agents are improving, and more people (independently of our contest) are discovering ways to use them to assist with learning and research. In some ways this is encouraging. But these same tools can still tend to overly flatter the user’s perspective, get overly fixated on particular lines of inquiry, summarise misleadingly or confabulate/hallucinate, fail to locate relevant content… Perhaps most importantly, few tools or systems are yet conducive to common knowledge.[2] Of course, in the important domains we most want to elevate — understanding frontier technology and science, and navigating high-stakes debates around power and politics — we don’t expect (or seek) unanimous agreement. But as we’ve previously written, enabling more people to have a fuller and more grounded picture of the overall conversation is both possible and desirable. Strong entries often exhibited four features: * Evaluation: many grand theories about AI for this and that turn out overblown or medi