Machine learning · AI safety
On Problems That Will Outlive Our Models
Which difficulties in machine learning safety would survive a complete change of methods?
Ideas worth
taking slowly.
Long-form notes on artificial intelligence, reasoning, and computation.
Essays & notes
02 essaysArtificial intelligence · Agent systems
On the Evaluation of Agent Harnesses
Separating what the model contributes to an agent from what the system around it contributes.
Elsewhere: GitHub · Google Scholar · X · Topics