PinnedNahom Birhan·Oct 17My Language Model Saw ‘jumped’ Seven Times and Knew What Came NextA next-token model with every gradient written by hand, and why 580,000 numbers out-predict a 20-million-cell table
PinnedNahom Birhan·Sep 14The 1990s Method That Beat My Neural NetworkWord embeddings from scratch, two ways — and the one with no training loop wins
PinnedNahom Birhan·Sep 6The Textbook Fix That Made My Language Model 5× WorseBuilding an n-gram language model from scratch and seeing what smoothing actually doesA response icon1A response icon1