Research
Thinking about how powerful systems shape society, and how we might build better guardrails and tools for the public.
This page is a living hub for active research projects, ideas-in-progress, and related writing.
Current work
Under what conditions do audit probability, detection accuracy, and penalty size yield stable compliance among strategic actors?
Designing audit and penalty regimes that sustain stable compliance in compute governance
in progress
Do Safety Constraints Impose Competitive Costs? Evidence from AI Model Refusal behaviour
Testing whether stricter moderation regimes impose measurable competitive costs in user-facing AI markets.
in progress
Writing & notes
- Longer essays and research-adjacent writing live on Powerlines .