Locking LLMs into blockchains as unbribeable, implacable judges could give us adjudication systems that are transparent, credibly neutral, and genuinely hard to game.
Prediction markets are the natural test case, and if we get it right, the implications will extend to any setting where judgment calls are required.
The core idea: at contract creation, you commit the exact model version and prompt on-chain. Everyone can inspect the full resolution mechanism before they trade. No rule changes mid-flight, no backroom negotiations, no discretionary judgment calls.
Why this helps:
--You can't bribe a model or flip its vote after the fact — the weights are fixed
--The LLM has no financial stake in the outcome, so conflicts of interest disappear
--The entire mechanism is auditable before anyone places a bet
It’s not a magic solution: models make mistakes, prompt design matters enormously, and information sources can still be targeted by adversaries. But these problems may be more tractable than the ones we're stuck with now---human bias, opacity, and the ever-present temptation to game committee decisions.
How to move forward:
--Experiment on lower-stakes contracts to build a track record and discover failure modes
--Standardize the approach as best practices emerge; help liquidity concentrate in markets with the most reliable LLM judges
--Build transparency tools so traders can actually inspect the model, prompt, and sources before trading
--Design ongoing human governance for meta-level decisions: which models to trust, when to update defaults, how to handle appeals, etc.
My new piece for
@a16zcrypto explores this in depth. With many thanks to
@sreeramkannan @ben_chain @benfielding and many others thinking about this area.
Link is below.