Personal Site

Henry Hu — I pull personalities out of language models and teach machines to play games.

Second-year CS at Northeastern, into interpretability and agent systems. I build things, take them apart to see how they work, and write up what's interesting here.

Selected work

All work

Dec 2025 - Present

Self-initiated project

Catan Learning Environment

An agent harness for LLM play in Settlers of Catan, built around a pure-Python engine and multi-agent training interface.

PythonAgentsMulti-agent RLGame engines

May 2026 - Present

Self-initiated research project

Activation Steering for Psychometric Trait Facets

A study of whether behavior-conditioned steering vectors for personality facets reflect the structure of human psychometric categories.

InterpretabilityActivation steeringQwenrepeng

Mar 2026 - May 2026

Software Engineer Intern

Doppel Labs / Wallace

Built core product surfaces for a research workspace: graph canvas, PDF reading, paper chat, and browser capture.

Next.jsYjsPDF UIBrowser extensions

From the journal

All posts

Currently

Working on backend systems and quietly obsessed with the parts of software that only get noticed when they break. Reading, playing more strategy games than I should, and slowly putting things into the journal here as they crystallize.