AI's real impact on development teams

Measuring and Improving Developer Experience πŸ“Š β€” with Abi Noda

Topic 7 of 7

Published Apr 17, 2025 Β· Clip 35:25–42:23 (6:57)

Abi discusses measuring AI's impact on development: DX data shows roughly 5-10% average productivity gains, best measured via self-reported time savings; PR throughput lifts on social media are mostly correlation, not causation. He raises a real concern: code maintainability scores are lower for teams using GenAI more, since tools excel at 0-to-1 boilerplate but most development is maintenance (1 to 1.1). He cites the contrast between LLMs acing abstract LeetCode (~85-90%) versus under 20% success on real tickets in real codebases, and warns executive expectations are inflated by hype. Beyond coding, he sees AI as well suited to the knowledge management problem via search/discoverability and AI-generated documentation.