DSPy, from Stanford NLP: define task signatures and let optimizers tune prompts
stanfordnlp · x · 2026-09-26
DSPy, the framework that started at Stanford NLP, pushes a "program, don't prompt" workflow:
- Describe a task as typed inputs and outputs (a signature) instead of hand-writing prompts
- Its optimizers automatically tune the prompts against a metric you choose
Rather than iteratively tweaking prompt text, developers describe tasks like ordinary programs and let DSPy handle prompt engineering.
More from coding & agent
- A four-dimension eval framework for long-horizon agents: outcome, trajectory, experience, governance — sh_reya · 2026-09-26
- Exa launches Agent Ultra, a swarm-of-agents deep research system it claims is SOTA — yoimnotkesku · 2026-09-26
- Databricks launches Unity Gateway CLI to centrally manage models across coding agents — matei_zaharia · 2026-09-26
- GitHub Cut Server Render Time 55% by Ditching CSS-in-JS, With Copilot Agents Finishing the Migration — DanWahlin · 2026-09-26
- Open-Source Lorca Launches: Native Mac Agent Client with iOS Pairing, Runs Agents Locally — dotey · 2026-09-26
- From libraries and starter apps to skills and lint rules: how packaging is changing — fernandorojo · 2026-09-26