Boris Cherny’s Viral Workflow Redefines Software Development, Signals New Era for AI Engineering

When the creator of the world’s most advanced coding agent speaks, Silicon Valley doesn’t just listen — it takes notes. For the past week, the global engineering community has been meticulously dissecting a viral thread on X from Boris Cherny, the visionary creator and head of Claude Code at Anthropic. What began as a seemingly casual sharing of his personal terminal setup has rapidly transformed into a de facto manifesto on the future of software development, with industry insiders hailing it as a pivotal, watershed moment for the burgeoning AI startup.

Cherny, a leading figure in the AI development space, revealed an unconventional yet remarkably effective workflow that leverages multiple AI agents in parallel, fundamentally altering the traditional coding paradigm. His insights quickly resonated across developer circles, prompting widespread discussion and adoption attempts. Jeff Tang, a prominent voice in the developer community known for his astute observations on technological shifts, encapsulated the sentiment perfectly: "If you’re not reading the Claude Code best practices straight from its creator, you’re behind as a programmer." The ripple effect extended further, with industry observer Kyle McNease declaring that Cherny’s "game-changing updates" have Anthropic "on fire," suggesting the company could be on the precipice of "their ChatGPT moment" — a reference to OpenAI’s breakthrough impact on the AI landscape.

The profound excitement surrounding Cherny’s revelation stems from a potent paradox: his workflow is surprisingly simple in its underlying principles, yet it empowers a single human operator to achieve the output capacity typically associated with a small, dedicated engineering department. This exponential leap in productivity caught the attention of many, with one user on X noting that after implementing Cherny’s setup, the experience "feels more like Starcraft" than conventional coding. This evocative analogy highlights a profound shift: developers are moving away from the laborious, line-by-line typing of syntax to a strategic role, commanding and orchestrating autonomous AI units to execute complex tasks.

This innovative approach is reshaping how software is built, offering a glimpse into a future where human ingenuity is amplified by intelligent automation. Here is a detailed analysis of the workflow, straight from the architect himself, illustrating how Anthropic is not just building AI but also redefining the very process of software creation.

How Running Five AI Agents at Once Turns Coding into a Real-Time Strategy Game

The most striking revelation from Cherny’s disclosure is his departure from the linear, sequential model of software development. In the traditional "inner loop" of development, a programmer typically focuses on a single task: writing a function, meticulously testing it, debugging, and only then moving to the next discrete component. Cherny, however, adopts a dramatically different posture, akin to a fleet commander overseeing multiple concurrent operations.

"I run 5 Claudes in parallel in my terminal," Cherny explained, detailing his setup. "I number my tabs 1-5, and use system notifications to know when a Claude needs input." This multi-agent orchestration transforms the development environment into a dynamic command center. By utilizing iTerm2 system notifications, Cherny efficiently manages five simultaneous work streams, each powered by a dedicated AI agent. While one Claude agent might be diligently running a comprehensive test suite to validate recent changes, another could be actively refactoring a legacy module, streamlining its architecture and improving maintainability. Concurrently, a third agent might be drafting detailed documentation for a new feature, ensuring clarity and completeness from the outset. This parallel processing eliminates idle time, allowing Cherny to fluidly switch between tasks that require his strategic oversight, rather than being bogged down in sequential execution.

Beyond his local terminal, Cherny further expands his operational capacity by running "5-10 Claudes on claude.ai" in his browser. This web-based interface provides additional flexibility, and a clever "teleport" command allows him to seamlessly hand off sessions between his web environment and his local machine, ensuring continuity and maximum efficiency. This sophisticated orchestration of AI agents directly validates the "do more with less" strategy articulated by Anthropic President Daniela Amodei earlier this week. While competing AI giants like OpenAI pursue massive, trillion-dollar infrastructure build-outs centered on raw computational power, Anthropic is demonstrating that superior orchestration, intelligent workflow design, and strategic utilization of existing models can yield exponential productivity gains, fundamentally challenging prevailing industry wisdom.

The Counterintuitive Case for Choosing the Slowest, Smartest Model

In an industry perpetually obsessed with minimizing latency and maximizing speed, Cherny’s choice of AI model came as a surprising, yet deeply insightful, revelation. He disclosed that he exclusively uses Anthropic’s heaviest, most powerful, and consequently, slowest model: Opus 4.5.

"I use Opus 4.5 with thinking for everything," Cherny explained. "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 counterintuitive decision highlights a critical understanding of the true bottlenecks in modern AI-assisted development. The primary constraint isn’t the raw speed at which an AI generates a token; rather, it is the significant human time and effort spent correcting the AI’s inevitable mistakes or guiding it through complex logic.

For enterprise technology leaders and developers alike, this is a profoundly critical insight. Cherny’s workflow suggests that strategically paying the "compute tax" for a more intelligent, albeit slower, model upfront ultimately eliminates the far more costly and time-consuming "correction tax" later in the development cycle. A smarter model, capable of more complex reasoning and superior tool use, requires less human intervention, fewer iterative corrections, and ultimately leads to a faster overall development velocity, despite its individual response time.

One Shared File Turns Every AI Mistake into a Permanent Lesson

Cherny also detailed an ingenious solution to a pervasive problem in AI development: "AI amnesia." Standard large language models typically do not "remember" a company’s specific coding style, architectural decisions, or nuanced preferences from one session to the next. This lack of persistent memory can lead to repetitive errors and a constant need for human re-instruction.

To address this, Cherny’s team maintains a single, crucial file named CLAUDE.md, meticulously checked into 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 wrote. This simple yet powerful practice transforms the codebase itself into a self-correcting, continuously learning organism. When a human developer reviews a pull request and identifies an error or a deviation from established guidelines, they don’t just fix the code; they also tag the AI by updating its instructions within CLAUDE.md. As Aakash Gupta, a product leader analyzing the thread, aptly noted, "Every mistake becomes a rule." This mechanism ensures that the AI’s learning is cumulative and shared across the team. The longer the team collaborates and contributes to this shared knowledge base, the smarter and more aligned the AI agent becomes, drastically reducing the incidence of repeated errors and fostering a continuously improving development cycle.

Slash Commands and Subagents Automate the Most Tedious Parts of Development

The "vanilla" workflow that one observer praised, despite its sophisticated underlying mechanics, is powered by rigorous automation of repetitive and often tedious tasks. Cherny extensively utilizes slash commands — custom shortcuts and scripts checked into the project’s repository — to handle complex operations with remarkable efficiency, often with a single keystroke.

He highlighted a command called **/commit-push-pr**, which he invokes dozens of times daily. Instead of manually typing a series of git commands, crafting a detailed commit message, and then navigating to a web interface to open a pull request, the AI agent autonomously handles this entire bureaucratic sequence of version control. This significantly reduces cognitive load and allows the developer to maintain focus on higher-level problem-solving.

Furthermore, Cherny deploys specialized "subagents" — AI personas tailored for specific phases of the development lifecycle. For instance, he might use a dedicated "code-simplifier" agent to refactor and clean up architectural complexities after the main development work is completed, ensuring code elegance and maintainability. Another "verify-app" agent could be tasked with running comprehensive end-to-end tests before anything is shipped, acting as a final quality gatekeeper. These specialized agents streamline and automate critical, often overlooked, aspects of the development process, ensuring robust and high-quality software delivery.

Why Verification Loops Are the Real Unlock for AI-Generated Code

If there is a single, overarching reason why Claude Code has reportedly achieved the significant milestone of $1 billion in annual recurring revenue so rapidly, it is likely its sophisticated implementation of the "verification loop." The AI in Cherny’s workflow is not merely a text generator or a code writer; it functions as an integrated tester and validator, capable of verifying its own work.

"Claude tests every single change I land to claude.ai/code using the Claude Chrome extension," Cherny wrote. "It opens a browser, tests the UI, and iterates until the code works and the UX feels good." This capability represents a monumental shift in AI-assisted development. Cherny argues that empowering the AI with the means to verify its own output — whether through browser automation for UI testing, executing bash commands for system-level checks, or running comprehensive test suites for code validation — improves the quality of the final result by an astounding "2-3x." The agent doesn’t just produce code; it proactively proves that the code functions as intended, significantly reducing the burden on human developers for quality assurance and error detection. This self-verification mechanism accelerates development cycles while simultaneously elevating the reliability and robustness of the generated software.

What Cherny’s Workflow Signals About the Future of Software Engineering

The overwhelming reaction to Cherny’s viral thread suggests a pivotal and irreversible shift in how developers perceive and approach their craft. For years, the concept of "AI coding" primarily meant an autocomplete function within a text editor — a helpful but ultimately superficial tool designed to speed up typing. Cherny’s demonstrated workflow, however, transcends this limited view, illustrating that AI can now function as an operating system for labor itself, fundamentally transforming the nature of human-computer interaction in development.

"Read this if you’re already an engineer… and want more power," Jeff Tang succinctly summarized on X. His statement underscores the profound implications: the tools to multiply human output by a factor of five or more are not a distant dream; they are already here, accessible, and demonstrably effective.

This paradigm shift demands more than just embracing new tools; it requires a fundamental recalibration of mindset. Developers must be willing to stop viewing AI as a mere assistant, a glorified autocomplete feature, and instead start treating it as an intelligent, autonomous workforce capable of executing complex tasks under strategic human command. The programmers who make this crucial mental leap first will not only experience unprecedented levels of productivity and efficiency but will also find themselves playing an entirely different game of software development — a real-time strategy game where human intellect orchestrates a powerful AI army. Everyone else, clinging to traditional methods, will still be typing, potentially finding themselves left far behind in the wake of this revolutionary transformation.