When Boris Cherny, the mastermind behind Claude Code—the world’s most advanced AI coding assistant—shares insights about his workflow, the tech community listens intently. Over the past week, a viral thread by Cherny on X (formerly Twitter) has sparked widespread discussion among developers and AI experts, who see his approach as a transformative shift in software engineering.
Cherny’s detailed revelation breaks from traditional linear coding methods and introduces a dynamic system where multiple AI agents operate simultaneously. This new paradigm has been hailed by industry voices as potentially marking Anthropic’s “ChatGPT moment,” a pivotal breakthrough in AI-powered development productivity.
Managing Five AI Agents Simultaneously: Coding as a Real-Time Strategy Game
Unlike conventional coding, where programmers write and test code sequentially, Cherny compares his workflow to commanding a fleet in a strategy game. He runs five Claude AI agents in parallel within his terminal, each dedicated to different tasks such as testing, refactoring, and documentation.
Using iTerm2 system notifications, Cherny effectively coordinates these concurrent streams, optimizing workflow efficiency. Additionally, he manages multiple Claude sessions through his browser, seamlessly teleporting between local and web environments. This multi-agent orchestration exemplifies Anthropic’s “do more with less” philosophy, contrasting with competitors like OpenAI who focus heavily on infrastructure expansion.
Choosing the Smartest, Not Fastest, AI Model for Better Results
In an industry often obsessed with latency, Cherny’s choice to use Anthropic’s largest and slowest AI model, Opus 4.5, is counterintuitive yet insightful. Despite its slower speed, Opus 4.5 requires less human intervention due to its superior understanding and tool usage, ultimately accelerating development by reducing error corrections.
This approach highlights a critical insight for enterprise AI adoption: investing in smarter AI upfront minimizes costly human corrections later, improving overall project turnaround times.
Transforming AI Errors into Learning Opportunities via a Shared File
A key challenge in AI-assisted coding is the models’ inability to retain organizational context across sessions. Cherny’s team addresses this by maintaining a single file, CLAUDE.md, within their git repository. Every AI misstep is documented here, teaching the AI to avoid repeating mistakes in future interactions.
This practice effectively turns the codebase into a self-improving system, where human reviewers not only fix errors but also update AI instructions, continuously refining the agent’s capabilities.
Slash Commands and Subagents Automate Repetitive Tasks
Cherny’s workflow leverages automation through slash commands—custom shortcuts integrated into the project repository—that streamline complex operations with minimal input. For example, the /commit-push-pr command handles version control tasks autonomously, eliminating manual steps.
He also employs specialized AI subagents responsible for distinct development phases, such as simplifying code architecture and conducting end-to-end application verification prior to deployment.
Verification Loops: The Key to Reliable AI-Generated Code
The rapid growth of Claude Code to over $1 billion in annual recurring revenue is attributed largely to its robust verification loop. Unlike simple text generators, Claude actively tests every code change through browser automation and test suites until the output meets quality and usability standards.
This self-verification boosts code reliability by two to three times, ensuring that the AI-generated solutions are not only syntactically correct but also functionally sound.
Implications for the Future of Software Engineering
Cherny’s workflow signals a profound shift in software development philosophy. No longer is AI merely an autocomplete helper; it is emerging as a comprehensive operational system that multiplies human productivity manifold.
As Jeff Tang, a respected developer commentator, summarized, “Read this if you’re already an engineer… and want more power.” The developers who adopt this mindset—treating AI as a workforce rather than just an assistant—will fundamentally change the coding landscape, moving from manual typing to orchestrated AI-driven creation.
This evolution suggests that the future of programming lies not in writing every line by hand but in commanding intelligent agents that collaborate autonomously, dramatically accelerating innovation and efficiency.
Fonte: ver artigo original
Related Articles
AWS AI Coding Tool Reportedly Caused 13-Hour Outage by Deleting Customer System
Google Leverages 5 Million News Articles and Gemini AI to Predict Flash Floods
Anthropic Unveils Cowork: A User-Friendly AI Desktop Agent to Manage Your Files Without Coding
Meta Secures 1 Gigawatt of Solar Power in Major Sustainability Push
