Creator of Claude Code Reveals His Workflow, Stunning Developers Worldwide

United States - Ekhbary News Agency

Creator of Claude Code Reveals His Workflow, Stunning Developers Worldwide

When the architect of the world's most sophisticated coding agent speaks, Silicon Valley doesn't just listen—it takes meticulous notes. For the past week, the engineering community has been captivated by a series of posts on X (formerly Twitter) from Boris Cherny, the visionary creator and lead of Claude Code at Anthropic. What began as a humble sharing of his personal terminal setup has rapidly evolved into a viral manifesto that is fundamentally reshaping perceptions of software development, with industry insiders hailing it as a pivotal moment for the AI startup.

Esteemed figures in the developer sphere have lauded Cherny's insights. Jeff Tang, a prominent voice in the community, remarked, "If you're not absorbing the Claude Code best practices directly from its creator, you're falling behind as a programmer." Kyle McNease, another keen observer of the industry, elevated the praise, declaring that with Cherny's "game-changing updates," Anthropic is "on fire" and potentially on the cusp of its own "ChatGPT moment."

The widespread excitement is rooted in a compelling paradox: Cherny's methodology is remarkably straightforward, yet it empowers a single individual to operate with the productivity capacity of a small, dedicated engineering team. One user, after implementing Cherny's setup, shared on X that the experience felt "more like Starcraft" than conventional coding, signifying a paradigm shift from manual syntax input to the strategic orchestration of autonomous digital entities.

This article delves into the workflow that is redefining software creation, directly from its originator. It explores how the concurrent operation of multiple AI agents transforms the coding process into an engaging real-time strategy game.

Beyond Linearity: Orchestrating Like a Fleet Commander

Perhaps the most striking revelation from Cherny's disclosure is his departure from linear coding practices. Traditional development, often termed the "inner loop," involves a programmer writing a function, testing it, and then proceeding to the next task. Cherny, however, adopts the posture of a fleet commander, orchestrating multiple AI agents simultaneously.

"I run 5 Claudes in parallel in my terminal," Cherny explained in his posts. "I number my tabs 1-5, and use system notifications to know when a Claude needs input."

By leveraging iTerm2's system notification capabilities, Cherny efficiently manages five distinct work streams concurrently. For instance, one AI agent might be executing a test suite, another could be refactoring a legacy module, while a third diligently drafts documentation. Furthermore, he extends this parallel processing to the web interface, running "5-10 Claudes on claude.ai" in his browser and utilizing a "teleport" command to seamlessly transfer sessions between the web and his local machine.

This sophisticated approach strongly validates the "do more with less" strategy recently articulated by Anthropic President Daniela Amodei. While major competitors like OpenAI are investing heavily in massive infrastructure build-outs, Anthropic is demonstrating that superior orchestration and intelligent utilization of existing AI models can unlock exponential gains in productivity.

The Counterintuitive Choice: Embracing the Slowest, Smartest Model

In a move that defies the industry's common obsession with low latency, Cherny revealed his exclusive reliance on Anthropic's most powerful, albeit slowest, model: Opus 4.5.

"I use Opus 4.5 with thinking for everything," Cherny elaborated. "It's the best coding model I've ever used, and even though it's bigger & slower than Sonnet, since you have to steer it less and it's better at tool use, it is almost always faster than using a smaller model in the end."

This insight is particularly critical for enterprise technology leaders. The primary bottleneck in contemporary AI development is not the speed of token generation; rather, it is the significant human time invested in correcting AI-generated errors. Cherny's workflow suggests that incurring the "compute tax" for a more intelligent model upfront can substantially mitigate the "correction tax" downstream.

A Shared File: Turning AI Mistakes into Permanent Lessons

Cherny also shed light on how his team tackles the persistent issue of AI "amnesia." Standard large language models often fail to retain a company's specific coding conventions or architectural decisions across different sessions.

To circumvent this, Cherny's team maintains a singular file, designated CLAUDE.md, within their Git repository. "Anytime we see Claude do something incorrectly we add it to the CLAUDE.md, so Claude knows not to do it next time," he stated.

This practice effectively transforms the codebase into a self-correcting organism. When a human developer reviews a pull request and identifies an error, their action extends beyond merely fixing the code; they instruct the AI to update its internal guidelines. "Every mistake becomes a rule," observed Aakash Gupta, a product leader who analyzed the thread. Consequently, the longer the team collaborates, the more adept and intelligent the AI agent becomes.

Slash Commands and Sub-Agents: Automating Tedious Development Tasks

The seemingly "vanilla" workflow that garnered praise is underpinned by rigorous automation of repetitive tasks. Cherny employs slash commands—custom shortcuts integrated into the project's repository—to execute complex operations with a single keystroke.

He highlighted a specific command, /commit-push-pr, which he invokes dozens of times daily. This automation eliminates the need for manual Git commands, commit message writing, and pull request initiation, allowing the AI to autonomously manage the bureaucratic aspects of version control.

Cherny further utilizes sub-agents, specialized AI personas, to manage distinct phases of the development lifecycle. He employs a code-simplifier agent to refine architecture post-development and a verify-app agent to conduct end-to-end testing before any deployment.

Verification Loops: The True Enabler of AI-Generated Code

If there's one singular reason Claude Code has reportedly achieved $1 billion in annual recurring revenue so rapidly, it is likely the implementation of robust verification loops. The AI functions not merely as a code generator but as an integrated testing system.

"Claude tests every single change I land to claude.ai/code using the Claude Chrome extension," Cherny revealed. "It opens a browser, tests the UI, and iterates until the code works and the UX feels good."

He posits that equipping the AI with the capability to verify its own output—whether through browser automation, executing bash commands, or running test suites—enhances the quality of the final product by a factor of "2-3x." The agent's role transcends code writing; it actively validates the code's functionality and user experience.

Cherny's Workflow Signals a New Era in Software Engineering

The fervent reaction to Cherny's thread suggests a fundamental paradigm shift in how developers perceive and practice their craft. For years, "AI coding" was largely synonymous with an enhanced autocomplete feature within a text editor—a convenient way to accelerate typing. Cherny has demonstrably proven that AI can now serve as a comprehensive operating system for labor itself.

"Read this if you're already an engineer... and want more power," Jeff Tang succinctly summarized on X, capturing the sentiment of empowerment.

The tools capable of multiplying human output by a factor of five are no longer theoretical; they are available now. Their effective implementation requires a simple yet profound mental leap: shifting the perspective from viewing AI as a mere assistant to embracing it as a scalable workforce. The programmers who make this cognitive transition first will not only achieve unprecedented productivity but will also find themselves engaged in an entirely different, more strategic, and ultimately more rewarding game of software development. The rest will continue with the traditional approach.

Ekhbary News Agency