Anthropic’s recent two-day event, Code with Claude, held in London, showcased transformative advancements in software development, coinciding with Google’s I/O. During the event, Anthropic engineer Jeremy Hadfield posed a thought-provoking question to attendees: “Who here has shipped a pull request entirely written by Claude?” To his surprise, nearly half the audience raised their hands, illustrating the growing reliance on AI-driven tools for coding tasks. Pull requests, essential for updating existing software, have traditionally consumed a significant portion of developers’ time. However, with Claude’s capabilities, many are now experiencing a shift in how they approach coding.
Anthropic’s Claude, alongside similar tools from competitors like OpenAI and Microsoft, is reshaping the software development landscape. Hadfield revealed that a substantial portion of the code at Anthropic is now generated by Claude, signaling a broader trend where top tech companies boast about minimal manual coding. The event highlighted the rapid acceptance of this new paradigm, as developers increasingly delegate coding responsibilities to AI. With recent updates, including features like “dreaming,” Claude Code is designed to autonomously refine its code, learning from past tasks to optimize future performance. As Boris Cherny, head of Claude Code, noted, the goal is to allow Claude to self-correct and manage errors without heavy human intervention.
Despite the enthusiasm at Code with Claude, concerns linger in the wider developer community. Many coders express skepticism about the over-reliance on AI, citing issues such as code quality and security vulnerabilities. Critics argue that AI-generated code often requires extensive review, counteracting the productivity gains management seeks. Conversations on platforms like Reddit and Hacker News highlight fears that developers may lose their coding skills as they become more dependent on AI. In response, Anthropic’s leadership emphasizes the importance of maintaining traditional software development practices, asserting that human oversight remains critical. As Katelyn Lesse, engineering lead at Anthropic, pointed out, while Claude’s capabilities are impressive, expert engineers are still essential for designing complex systems and troubleshooting. The journey toward fully automated coding continues, but balancing innovation with established best practices remains crucial.
Source: Anthropic’s Code with Claude showed off coding’s future—whether you like it or not via MIT Technology Review
