Boris Cherny's Innovative Workflow Revolutionises Software Development Practices
Boris Cherny, the head of Claude Code at Anthropic, has sparked significant interest among software engineers with a recent thread on X. This discussion, which began with a glimpse into his personal terminal setup, has evolved into a transformative manifesto on the future of coding, with many within the industry viewing it as a landmark moment.
Prominent figures in the developer community have emphasized the importance of Cherny's insights. “If you’re not reading the Claude Code best practices straight from its creator, you’re behind as a programmer,” stated Jeff Tang, a notable voice in tech circles. Another expert, Kyle McNease, highlighted Cherny’s “game-changing updates,” suggesting that Anthropic is on the brink of a pivotal breakthrough in AI technology.
The buzz around Cherny's workflow stems from its unexpected simplicity, which allows one individual to achieve an output comparable to that of a small engineering team. Users on X remarked that implementing Cherny's setup felt more akin to a strategy game than conventional coding, as it shifts the focus from typing to directing complex systems.
Cherny's most striking revelation is his non-linear coding process, which contrasts sharply with the traditional development method. Instead of progressing through tasks sequentially, he likens himself to a fleet commander. “I run 5 Claudes in parallel in my terminal,” Cherny explained. He employs system notifications to manage five simultaneous tasks, including testing, legacy refactoring, and documentation drafting, all while also utilizing multiple Claudes in his browser.
This innovative approach underscores the “do more with less” strategy previously articulated by Anthropic’s President, Daniela Amodei. While competitors focus on significant infrastructure expansion, Anthropic is proving that effectively orchestrating existing AI models can markedly enhance productivity.
In an unexpected twist for an industry preoccupied with speed, Cherny exclusively employs the slowest and most comprehensive AI model, Opus 4.5. “I use Opus 4.5 with thinking for everything,” Cherny noted. He posited that, despite its size and slower speed, its superior problem-solving capabilities make it more efficient in the long run.
This insight is crucial for enterprise technology leaders, as it reveals that the bottleneck in artificial intelligence development lies not in quick token generation but in the time spent correcting mistakes made by AI. By opting for a smarter model, developers can mitigate errors upfront, saving time later.
Addressing AI’s challenge of forgetfulness, Cherny shared how his team has tackled the issue through meticulous documentation. They maintain a single file named CLAUDE.md, in which they log every mistake made by the AI. According to Cherny, “Anytime we see Claude do something incorrectly we add it to the CLAUDE.md, so Claude knows not to do it next time.” This ensures that the AI improves continuously with each interaction.
Cherny's workflow is also bolstered by rigorous automation of repetitive tasks using slash commands. He frequently relies on commands such as /commit-push-pr to streamline version control processes, thus saving time and effort.
One of the standout features of Claude Code is its verification loop. Cherny emphasizes the importance of this loop: “Claude tests every single change I land to claude.ai/code using the Claude Chrome extension.” This capability allows the AI to verify its own code, significantly improving the efficiency and quality of the software produced.
The enthusiastic response to Cherny’s workflow indicates a significant shift in the software engineering landscape. Traditionally, AI coding has been viewed as a mere autocomplete feature within text editors. However, Cherny demonstrates that AI can now serve as a complete operating system for software development.
“Read this if you’re already an engineer… and want more power,” summarized Jeff Tang on X. The tools necessary to amplify human output dramatically are already available, calling for a paradigm shift in how AI is perceived—not merely as an assistant but as an integral workforce in the development process.