# How Claude Code Claude Codes | The Vergecast Source: https://www.youtube.com/watch?v=IvDnGRFMmgE Recap page: https://rapidrecap.app/video/IvDnGRFMmgE Generated: 2026-02-24T13:33:29.527+00:00 --- ## Quick Overview Boris Churney, creator of Claude Code at Anthropic, revealed that his personal coding dependency shifted suddenly to 100% reliance on the AI with the release of Opus 4.5, marking a major step change where the AI handles testing and final fixes, eliminating the need for manual text editor adjustments. **Key Points:** - Boris Churney stated that Claude Code went from writing 10% of his code (Sonnet 3.5 new) to 100% following the launch of Opus 4.5 in November, a change that felt sudden and natural. - The shift to 100% reliance involved the AI testing code, verifying website functionality via a browser, and fixing minor issues like pixel misalignment without manual intervention from the engineer. - Anthropic believes the path to safe AGI involves increasing capabilities through coding, then tool use, and finally computer use, with coding always serving as the mediating layer. - Claude Code was initially designed as a developer product for developers, but non-developers quickly adopted it, leading Anthropic to launch Co-work for less technical users with built-in safety features like a virtual machine and deletion protection. - Co-work experienced immediate exponential growth upon launch, contrasting with Claude Code's slower initial adoption, driven by pent-up demand for a more understandable tool to automate busy work. - David Pierce advocates for using AI to automate busy work, citing examples like using Co-work to organize screenshots or pay parking tickets, rather than immediately jumping to building complex daily planners. - The optimal UI for these tools remains undiscovered, but Anthropic is constantly experimenting across various surfaces (terminal, IDEs, iOS app) and values proactive behavior, though getting the boundary right to avoid annoyance is challenging. **Context:** The Vergecast episode features an interview with Boris Churney, the creator of Claude Code at Anthropic, discussing the one-year anniversary of the AI coding product. The conversation centers on how Claude Code, particularly with recent model upgrades, has changed software development, moving from code generation to agents capable of tool use and computer interaction, while also exploring the expanding user base beyond traditional developers via products like Co-work. ## Detailed Analysis The discussion centers on the rapid evolution and impact of Claude Code, with its creator, Boris Churney, detailing his personal experience of moving from writing code to having Claude Code handle 100% of his coding tasks after the Opus 4.5 release, which enabled automated testing and self-correction of code, eliminating manual fiddling in text editors. Anthropic views coding as the fundamental step toward safe AGI, preceding tool use and general computer use. While initially a developer tool, the surprising adoption by non-developers prompted the creation of Co-work, a user-friendlier interface built on the same agent SDK but with added safety measures unsuitable for engineers who prefer maximum customization in the highly configurable core Claude Code. Churney noted that Co-work saw immediate exponential adoption because it addressed pent-up demand for automating busy work, such as organizing screenshots or managing project status updates, contrasting with Claude Code's slower initial uptake. Experts suggest users should evaluate AI tools with a sharper eye than traditional services due to novelty and rapid development, although Anthropic emphasizes enterprise focus, which necessitates strong security and privacy controls; for instance, Churney cannot view user data without explicit reproduction steps, and Co-work operates within a secure virtual machine boundary, though prompt injection remains an area requiring ongoing vigilance. ### Claude Code's Evolution - Shifted from writing 10% of Churney's code on Sonnet 3.5 new to 100% on Opus 4.5 - The key change involved Opus testing code and verifying functionality via browser interaction - Anthropic believes the development path is code, then tool use, then computer use. ### Developer Experience Transformation - Engineers are accustomed to tech stack changes, but moving away from working with source code is a big shift - The ability for the AI to test and fix code automatically means engineers can move to the next task faster. ### User Base Expansion Beyond Development - Initially designed for developers, non-developers started using Claude Code, leading to the creation of Co-work - Companies like Spotify and Shopify now have non-engineers (product managers, data scientists) using the tools. ### Co-work vs. Claude Code Adoption - Co-work saw immediate exponential success due to pent-up demand for a more understandable tool automating busy work - Examples of Co-work use include organizing screenshots, paying parking tickets, and managing team status updates via Slack. ### Security and Privacy Stance - Anthropic prioritizes safety, aligning with enterprise customers' needs for security and privacy - Churney noted he cannot view user data for bug reproduction due to strict controls, and Co-work runs in a virtual machine for isolation. ### Ongoing UI Exploration - The UI of the future is undiscovered, leading to constant experimentation across surfaces like IDE extensions, iOS apps, and desktop apps - The team monitors user feedback on Twitter and Threads to guide development, exploring proactivity while balancing the risk of being intrusive.