Over the past week, the engineering community has been analyzing a thread about X from Boris Cherny, creator and lead of Claude Code at Anthropic. What started as an informal share of his personal terminal setup turned into a viral statement about the future of software development, with industry insiders calling it a watershed moment for the startup.
“If you’re not reading Claude Code best practices straight from its creator, you’re falling behind as a programmer,” wrote Jeff Tang, a prominent voice in the developer community. Kyle McNeese, another industry observer, went further, declaring that with Cherny’s “game-changing updates,” Anthropic is “on fire,” and will likely experience “its own ChatGPT moment.”
The excitement stems from a paradox: Cherny’s workflow is surprisingly simple, but it allows a single human to work at the production capacity of a small engineering department. As one user noted on
Here’s an analysis of the workflow that’s reshaping how software is built, straight from the architect himself.
How running five AI agents simultaneously turns programming into a real-time strategy game
The most striking finding from Cherny’s revelation is that he does not encode in a linear manner. In the traditional “inner loop” of development, a programmer writes a function, tests it, and then moves on to the next function. But Cherny serves as the fleet commander.
“I’m running 5 clouds in parallel on my station,” Cherny wrote. “I number the tabs from 1 to 5, and use system notifications to know when Claude needs input.”
By using iTerm2 notifications, Cherny effectively manages five simultaneous workflows. While one agent runs a test suite, another agent rebuilds an old unit, and a third agent drafts documentation. He also runs “5-10 Claudes on claude.ai” in his browser, using the “teleport” command to hand off sessions between the web and his local machine.
This validates the “do more with less” strategy outlined by humanitarian chief Daniela Amodei earlier this week. While competitors like OpenAI seek to build trillion-dollar infrastructure, Anthropic is proving that superior coordination of existing models can lead to massive productivity gains.
The counterintuitive case of choosing the slowest and smartest model
In a surprising move for an industry obsessed with latency, Cherny revealed that he exclusively uses Anthropic’s heaviest and slowest model: the Opus 4.5.
“I use Opus 4.5 to think about everything,” Czerny explained. “It’s the best programming model I’ve ever used, and even though it’s bigger and slower than Sonnet, because you have to route it less and it’s better at using tools, it’s almost always faster than using a smaller model in the end.”
For enterprise technology leaders, this is a critical insight. The bottleneck in modern AI development is not the speed of token generation; It’s the time a human spends debugging AI. Cherny’s workflow suggests that paying the “computation tax” of a smarter model upfront eliminates the “correction tax” later.
One shared file turns every AI mistake into a lasting lesson
Cherny also detailed how his team solved the problem of AI amnesia. Standard large language models do not “remember” a company’s coding style or architectural decisions from one session to the next.
To address this issue, Cherny’s team maintains a single file called CLAUDE.md in their git repository. “Anytime we see Claude doing something incorrectly, we add it to CLAUDE.md, so Claude knows he won’t do it next time,” he wrote.
This practice turns the code base into a self-correcting object. When a human developer reviews a pull request and discovers a bug, he doesn’t just fix the code; They tag the AI to update its own instructions. “Every mistake becomes a rule,” noted Akash Gupta, a product lead analyzing the topic. The longer a team works together, the smarter the agent becomes.
Trim commands and subagents automate the most tedious parts of development
The “vanilla” workflow praised by one observer is supported by rigorous automation of repetitive tasks. Cherny uses slash commands-custom shortcuts verified in the project repository-to handle complex operations with a single keystroke.
He highlighted something called /commit-payment-prWhich he conjures dozens of times daily. Instead of manually writing git commands, writing a commit message, and opening a pull request, the agent handles the version control bureaucracy independently.
Cherny also deploys sub-agents – specialized AI personas – to handle specific stages of the development life cycle. It uses a code simplifier to clean up the build after the main work is done and an application validation agent to run extensive tests before shipping anything.
Why are validation loops the real solution for AI-generated code?
If there’s one reason why CloudCode has reached $1 billion in annual recurring revenue so quickly, it’s probably the Verification Loop. AI is not just a text generator; It’s a test.
“Every change I make to claude.ai/code is tested by Claude using the Claude Chrome extension,” Cherny wrote. “It opens the browser, tests the UI, and iterates until the code works and the user experience is good.”
He argues that giving AI a way to verify its work – whether through browser automation, running bash commands, or executing test suites – improves the quality of the end result by “2-3x.” An agent doesn’t just write code; Prove that the code works.
What Cherny’s workflow suggests about the future of software engineering
The reaction to Cherny’s thread indicates a pivotal shift in how developers think about their craft. For many years, “AI coding” meant a text editor’s autocomplete function, a faster way to type. Cherny has proven that it can now act as an operating system for the business itself.
“Read this if you’re already an engineer…and want more power,” Jeff Tang summed up on X.
The tools to multiply human output by a factor of five already exist. It just requires a willingness to stop thinking of AI as an assistant and start treating it as a workforce. Programmers who make this mental leap first will not only be more productive. They’ll be playing a completely different game, and everyone will still be writing.
