Essays on AI engineering, evaluation, agents and Python. Published on Medium, collected here.
Most AI projects don’t fail because the model is bad. They fail because the infrastructure around the model was picked backwards.