DSPy shows only 4.7M monthly downloads compared to LangChain's 222M, revealing puzzling underadoption. Companies using DSPy report benefits: faster model testing, more maintainable systems, and focus on context rather than plumbing. The framework solves major AI engineering challenges yet remains niche.

DSPy's problem isn't being wrong—it's being hard. Unfamiliar abstractions force teams to think differently when they just want pain relief. Khattab's Law now applies: any sufficiently complex AI system contains a buggy implementation of half of DSPy. Teams build these patterns anyway, just worse.

The evolution of every AI system follows a predictable path from simple API calls to complex RAG implementations with evaluations. Teams repeatedly reinvent DSPy's patterns through stages of adding retries, structured outputs, and evaluation frameworks.