Humboldt AI Team Wins $800 Prize for Lie-Detection Tool Targeting Hallucinations

Thomas Edrington walked into a room of roughly 200 programmers at Frontier Tower in San Francisco and walked out with an AI Safety prize for a tool his team calls a lie detector for artificial intelligence.

Edrington, executive director of the Transparent Humboldt Coalition and head of its affiliated Liberation Labs, led a team of local researchers and students to a first-place finish in the AI Safety category at the "Intelligence at the Frontier" hackathon, held as part of the Funding of the Commons conference. The team claimed an $800 prize, awarded specifically for the "meaningful advancement" of an existing project.

The project is called JiminAI Cricket, a name that is patent pending. The reference is deliberate: "a nod to Jiminy Cricket, Pinocchio's conscience: small but constantly present, ethical, and observant." The tool targets one of the most widely criticized failures of modern generative AI, the tendency to fabricate answers with total confidence. "When you ask an AI a question it doesn't know the answer to, it often makes one up, confidently, convincingly, but completely wrong," according to reporting on the project. In high-stakes fields like medicine, law, and finance, that flaw carries real consequences.

What sets JiminAI Cricket apart from after-the-fact fact-checking is its approach: rather than reviewing an AI's output once it is delivered, the tool monitors the process as it unfolds. It "watches an AI as it is thinking, and flags the thought process when something goes wrong, effectively identifying any detour in the answering process indicating that an answer is being crafted outside of the scope of provable facts."

The broader promise being attached to the win is significant. Edrington's team returned from San Francisco carrying not just an $800 check but what some observers have framed as a potential multi-million dollar industry breakthrough, though no commercialization timeline or technical validation has been publicly released. No further details on patent filing dates, model compatibility, or evaluation metrics have been disclosed.

Liberation Labs operates under the Transparent Humboldt Coalition, rooting what is now a nationally competitive AI safety project in Humboldt County's academic and research community.