Summary:
1. Boris Cherny, creator of Claude Code at Anthropic, has shared his revolutionary coding workflow that allows a single human to operate with the output capacity of a small engineering department.
2. Cherny’s approach involves running multiple AI agents simultaneously, using the slowest but smartest model, and maintaining a shared file to improve AI learning.
3. His automation of repetitive tasks, use of slash commands, and emphasis on verification loops signal a new era in software development where AI is not just an assistant but a workforce.
Article:
When Boris Cherny, the mastermind behind Claude Code at Anthropic, speaks about coding, Silicon Valley takes notes. His recent thread on X has sent shockwaves through the engineering community, sparking a discussion on the future of software development. Cherny’s innovative workflow challenges traditional coding methods, offering a glimpse into a new era where AI transforms from an assistant into a powerful workforce.
One of the key aspects of Cherny’s workflow is his ability to run multiple AI agents simultaneously, effectively turning coding into a real-time strategy game. By managing five Claudes in parallel in his terminal, Cherny maximizes productivity and efficiency, allowing for seamless orchestration of tasks. This approach aligns with Anthropic’s strategy of doing more with less, emphasizing superior orchestration over trillion-dollar infrastructure build-outs.
In a surprising twist, Cherny exclusively uses Anthropic’s slowest and smartest model, Opus 4.5, challenging the industry’s obsession with latency. His rationale for choosing a smarter model upfront is to eliminate the time spent correcting AI mistakes, ultimately leading to faster and more accurate results. Additionally, Cherny’s team maintains a shared file named CLAUDE.md in their git repository to ensure that AI learns from its errors, turning every mistake into a permanent lesson.
Automation plays a crucial role in Cherny’s workflow, with slash commands and subagents handling repetitive tasks and specific phases of the development lifecycle. By deploying these tools, Cherny streamlines operations and boosts efficiency, paving the way for a more productive coding experience. His emphasis on verification loops further enhances the quality of AI-generated code, ensuring that the final result meets rigorous standards.
Overall, Cherny’s workflow signals a paradigm shift in software engineering, urging developers to rethink their approach to coding. By treating AI as a workforce rather than a mere assistant, programmers can unlock new levels of productivity and efficiency. The tools to multiply human output are already here, waiting for those willing to embrace a new way of thinking. As Cherny’s thread continues to make waves in the industry, it’s clear that the future of software development is evolving, and those who adapt will lead the way in this new era of coding.
