artificial intelligence, its future trajectory, and its implications for humanity.
bookshelf
papers, books, and videos i found useful.
how attention residuals influence model behavior and internal representations.
efficient post-training quantization that preserves accuracy.
self-attention without a token’s own value vector.
a visual explanation of entropy, information, compression, and intelligence.
models improving by generating challenges and learning from the feedback loop.
nvidia’s unified world-model framework for language, media, and action.
a visual explanation of cross-entropy through compression, prediction, and probability.