Stanford's DSPy hits 39k GitHub stars for programming-not-prompting LLMs

_jaydeepkarale · x · 2026-11-16

DSPy, Stanford NLP's open-source framework for programming—rather than prompting—language models, has reached 39.1k GitHub stars and 3.5k forks across 4,865 commits. DSPy (Declarative Self-improving Python) lets developers build modular AI systems with compositional Python code instead of brittle prompts, and ships algorithms that automatically optimize both prompts and weights—applicable to classifiers, RAG pipelines, and agent loops. MIT-licensed, docs at dspy.ai.

Original post →

More from coding & agent

coding & agent channel →