Aleksey Komissarov

First-pass poisoning in neural solvers can drastically undermine solution accuracy, but clever data augmentation can flip the script, boosting performance from near-zero to nearly perfect.

AI literacy isn't just about knowing facts; it's about building cognitive antibodies through exposure to AI's flaws, like sycophantic validation and authority projection.