If DSPy is So Great, Why Isn't Anyone Using It?
Summary
The article critiques the DSPy pattern for AI engineering, noting a steep learning curve and a tendency to produce ad hoc versions. It outlines seven stages of AI pipeline evolution and presents two practical paths for teams: adopt DSPy with its patterns and a formal eval infrastructure, or implement the same ideas without the tool, focusing on typed interfaces, modular units, and early evaluation.