In this podcast episode, Alex from Anthropic hosts Lukas, Aman, and Jacob from Cursor to discuss how they are using Claude to build Cursor, focusing on the recursive feedback loop and the evolution of their product. They delve into the impact of model improvements, like 3.5 Sonnet, on Cursor's capabilities, particularly in multi-file editing and agentic coding. The conversation explores the advantages of being their own best customer, the spectrum of AI programming use cases within Cursor, and the potential of Background Agent for asynchronous tasks. They also address the challenges of code verification in large codebases, the importance of organizational knowledge, and the future of software development, including the role of AI-generated code and the evolving skills of software engineers. The group shares their positive experiences with Claude's new models, discussing the balance between coding and other areas of AI development, and speculate on the future of software engineering with AI.
