Now: AI safety research

Griffin Hilly

Finding my comparative advantage on an uncertain frontier.

Quant researcher. Amateur philosopher. Eternal student.
Husband to Madi Hilly. Father of two and counting. Proud Chicagoan.

Research

Thus far, my attempted mapping has covered 218 sources, producing 129 unique ideas across 11 category groups.

How far will models push us to be our best selves? Split by salience and stakes.

Putting human empowerment on a scale that can guide training.

Alignment by Onion: virtues in the model, rules in the harness, consequences in deployment.

Paul Collier’s 2007 book argued the poorest countries were stuck in development traps. Was it true, and is it still today?

An open-source prerequisite knowledge graph: more than 15,000 topics across 19 domains of human knowledge.

A philosophical framework built around agency — the capacity for, and tendency toward, rational autonomous will.

In progress

There's been little focus on the most tangible benefits of AI deployment: saving one billion years of human life annually.

In progress

Running dozens of Claude Code sessions from a phone: an always-on desktop, a written brief for each task, and a pipeline that files saved posts.