Founder of Minerval, building the claim graph. AI strategist at Broad Reach.
Highly compressed observations. Tweets, for someone who doesn't tweet.
An order-of-magnitude estimate of what it costs to build and serve the claim graph, doing it right. The cost is dominated by LLM inference; a frontier-quality build lands in the high single-digit millions, concentrated in the broad middle tiers and the ingestion firehose rather than the flagship claims.
Part I of A Marketplace of Ideas. How the prestige economy of science aligns incentives in some ways and fails badly in others.
Part II of A Marketplace of Ideas. What the scientific economy would look like if articles paid royalties roughly commensurate with their impact.
Part III of A Marketplace of Ideas. A new form of intellectual property that pays for an article's impact while leaving the science fully open.
Part IV of A Marketplace of Ideas. How to reward impact without recreating the perverse incentives, from review boards to impact-weighted citations and confidential vouchers.
Part V of A Marketplace of Ideas. How a single foundation could prove the concept, how government might follow, and what the scheme does not fix.
LLM memetics have so far proven tame, but multi-agent networks will change that. A sketch of how to test models for their propensity to spread memes in multi-agent setups.
Twelve testable properties an agentic claim graph should satisfy, from idempotency to calibration, with notes on how to measure each and where adversarial inputs make it harder.