Steve Yegge, Sourcegraph, ETLS Vegas (10)
summary
Claim: measuring token counts and AI percentage of code is outdated and completions are a gimmick, so throw away metrics like acceptance rates and token counting. Coding workflow shifts to chat-oriented programming where the interface is conversational and the model participates in design, debugging, and edits. Evidence comes from daily coders reporting faster change cycles and demos showing a working chat-driven dev loop rather than isolated code completions. Practical examples and failure modes are described.
AI-written summary · about
related talks
32mFrom Vibes to Outages: Riding the AI Code Wave
27mHow AI is changing Software Engineering: A Conversation with Gergely Orosz, @pragmaticengineer
27mThe Many Ends of Programming
43mFrom Clickers to Coders: The SDET Evolution Story
53mThe Human Side of AI: Unlocking Opportunity in the New Bottlenecks of Software Development
52mKeynote: The Psychology of Code
46mPanel Discussion: AI's Impact on Developers
46mPanel Discussion: The Agentic Software Factory That Enterprises Actually Need