When the person who built the most advanced coding agent on the market casually drops his terminal setup on X, people pay attention. Boris Cherny, creator and head of <a href="https://biz.allwinchina.org/ai-tools/claude-code/" title="Claude Code review”>Claude Code at Anthropic, did exactly that last week, and the engineering community has been picking apart his thread ever since.
What started as a seemingly mundane sharing of personal workflow spiraled into something bigger. Jeff Tang, a well-known developer advocate, put it bluntly: “If you’re not reading the Claude Code best practices straight from its creator, you’re behind as a programmer.” Kyle McNease went further, saying Anthropic is “on fire” and potentially facing “their ChatGPT moment.”
The irony is that Cherny’s workflow isn’t complicated. It’s actually surprisingly simple. But the output it enables is absurd: a single human operating with the throughput of a small engineering team. One user who tried implementing the setup described the experience as “feels more like Starcraft” than traditional coding. That’s not hyperbole.
Running five agents like a fleet commander
Cherny doesn’t code linearly. No writing a function, testing it, moving to the next. Instead, he acts like a fleet commander.
“I run 5 Claudes in parallel in my terminal,” he wrote. “I number my tabs 1-5, and use system notifications to know when a Claude needs input.”
Using iTerm2 system notifications, he manages five simultaneous work streams. One agent runs tests, another refactors a legacy module, a third drafts documentation. Meanwhile, he has 5-10 more Claudes running on claude.ai in his browser, with a “teleport” command to hand off sessions between web and local machine.
This validates the “do more with less” strategy Anthropic’s president Daniela Amodei talked about recently. While OpenAI is chasing trillion-dollar infrastructure, Anthropic is showing that better orchestration of existing models can yield exponential productivity gains. I’ve seen similar patterns with other tools, but the scale here is different.
The counterintuitive case for the slowest model
In an industry obsessed with latency, Cherny made a surprising choice: he uses only Anthropic’s heaviest, slowest model, Opus 4.5.
“I use Opus 4.5 with thinking for everything,” he 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 is a critical insight for anyone running AI-assisted development. The bottleneck isn’t token generation speed. It’s the human time spent correcting the AI’s mistakes. Paying the “compute tax” for a smarter model upfront eliminates the “correction tax” later. I’ve seen teams burn hours fixing output from fast-but-dumb models, and this approach directly addresses that.
One file that turns every mistake into a permanent lesson
Cherny also tackled the problem of AI amnesia. Standard LLMs don’t remember your coding style or architectural decisions between sessions.
His solution is a single file called CLAUDE.md stored in the 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 turns the codebase into a self-correcting organism. When a human reviews a pull request and spots an error, they don’t just fix the code. They tag the AI to update its own instructions. Every mistake becomes a rule. The longer the team works together, the smarter the agent gets. This is higher quality than I expected from a simple text file approach.
Slash commands and subagents automate the tedious stuff
The “vanilla” workflow people praised is powered by rigorous automation. Cherny uses slash commands—custom shortcuts checked into the project’s repository—to handle complex operations with a single keystroke.
He highlighted a command called /commit-push-pr, which he invokes dozens of times daily. Instead of manually typing git commands, writing a commit message, and opening a pull request, the agent handles everything. This approach has been tried before in various forms, but the integration with Claude Code makes it seamless.
What this means for the rest of us
Cherny’s workflow isn’t just about individual productivity. It signals a shift in how software development might work going forward. The role of the developer is moving from typing syntax to commanding autonomous units. That’s a fundamental change.
The real test will be whether this scales beyond a single genius and his team. But for now, the pattern is clear: better orchestration, smarter models, and institutional memory built into the codebase. I’m already experimenting with a CLAUDE.md file in my own projects, and the early results are promising.
Comments (0)
Login Log in to comment.
Be the first to comment!