DSPy: Programming the Objective, Not the Prompt
An intuitive introduction to building language-model systems whose prompts can be measured, optimized, and replaced.
AI
Notes on building, understanding, and learning with AI.
SIMBA: Learning Rules from Better and Worse Attempts
How DSPy’s stochastic introspective optimizer turns variable model behavior into rules and demonstrations.
GEPA: Learning from Explanations
An intuitive guide to reflective prompt evolution, Pareto candidate selection, and feedback-driven improvement of language-model programs.
How SWE-bench Measures AI Coding Agents
How SWE-bench turns real GitHub issues into executable evaluations—and what its grader can and cannot tell us.
When Code Is No Longer the Bottleneck
As AI makes code easier to produce, understanding, judgment, and ownership become the scarce resources in software engineering.