In this episode of Syntax, Wes interviews Kyle Cesmat from Coinbase about the company's use of AI in software development. Kyle explains that Coinbase is writing 40% of its code with AI, aiming for 50%, and discusses the tools and processes they use, including Cursor and Claude Code. He emphasizes that AI-generated code goes through a rigorous review process and that AI should be attributed and not make final decisions. The conversation covers the types of code being written with AI (primarily TypeScript and Go), the use of MCPs (Managed Component Protocol) for internal tooling, and the importance of code quality and security. Kyle also touches on the impact of AI on developer velocity, code review, and the potential for AI natives in the workforce, highlighting the balance between AI assistance and human oversight in software engineering.
Sign in to continue reading, translating and more.
Continue