The Unreasonable Effectiveness of Separating the Task from the Model — Maxime Rivest, DSPy
DSPy brings software engineering function principles to AI programs by separating task from model
“If for your repeated AI task you define an input interface and an output interface, you get to play in the internals. You get a lot of agility.”
DSPy co-creator Maxime Rivest argues that AI workflows should be structured like software functions — with defined inputs, outputs, and swappable internals — to gain reusability, composability, and model-agnosticism. The core insight is that separating the task contract from the model/prompt implementation lets teams swap models or techniques without rewriting logic. This is a practitioner-level engineering discipline talk rather than a major announcement, but directly relevant to teams building production AI systems.