Code
For a more up-to-date list of what I'm working on, visit my GitHub.

Systematic Sandbagging Evaluations on Claude 3.5 Sonnet
An investigation into Claude 3.5 Sonnet's sandbagging behaviors, examining how different consequences affect model performance across various tasks. This study extends previous research on AI systems' intentional underperformance.

micrograd-go
A Go implementation of Andrej Karpathy's micrograd, implementing backpropagation over a DAG of Value operations.