Essays on AI governance and on what a decade of crowdsourcing and consensus research teaches about trusting intelligent systems.
Over the last ten weeks I logged 148 Claude Code sessions across 12 projects: 86,838 events, 832 subagents, 2.2 billion tokens. I built a local analyzer that reads session logs, reconstructs every API call, and prices every token at the API list rate in effect the day it ran.
AI-assisted writing needs transparency: authors should measure and disclose how much an LLM contributed to their text. This essay introduces Refine, a Notion integration that tracks and scores AI involvement paragraph by paragraph.
A token only survives if it is treated as a product people want to buy, use, and hold. This essay applies churn analysis (subgroup discovery and behavioral timeline mining) to find out who is dumping a token and why.
Open on-chain voting exposes Cosmos governance to social pressure, bribery, and collusion. Fully homomorphic encryption can make votes secret while keeping tallies verifiable. This essay outlines how.
Elsewhere
I also occasionally write on the Authrty blog and on the Sydyk blog.