Writing archive
Earlier essays.
These pieces are older and exploratory in nature, touching on model security, reward hacking, red-team incentives, biosecurity governance, and how technical systems shape real-world risk.
AGI Model Weight Theft is Realistic
Why model weights should be treated as high-value infrastructure and not just files.
Why Reward Hacking is More Dangerous Than it Seems
A bridge between technical failure modes and the systems that incentivize them.
Building a Backbone for Benchtop Synthesizer Governance
Governance thinking at the boundary of cyber, biosecurity, and emerging tooling.
Managing Expectations in a Self-Echoic World
How LLM encouragement can distort expectations in real-world, high-stakes situations.
The Dual-Use Nature of AI Red Teaming Competitions
How adversarial discovery can improve safeguards while also creating information hazards.