Are LLMs not getting better?
Summary
The article questions whether LLMs have actually improved in programming tasks, using data and Brier scores to argue that progress has stalled since 2024-2025. It challenges readers to reconsider assumed AI advancement and discusses implications for practitioners.