Quick Answer: Boris Cherny, creator of Claude Code at Anthropic, hasn't written code by hand in over two months. He ships 22-27 pull requests daily—each one 100% written by Claude. Company-wide, Anthropic reports 70-90% of their code is AI-generated. This isn't a vision of the future; it's happening right now at one of the world's leading AI companies.
When the creator of one of the most popular AI coding tools reveals he no longer writes code himself, it's worth paying attention. Boris Cherny's announcement in January 2026 sent shockwaves through the developer community—and for good reason.
Here's exactly how he does it, what it means for developers everywhere, and how you can start adopting a similar workflow.
Watch: Claude Code Tutorial - From Zero to Building Features
Peter Yang's beginner tutorial demonstrating the same workflow Boris Cherny uses: Plan mode, CLAUDE.md, and to-do lists before coding.
The Shocking Revelation: 22-27 PRs Daily
In a post that went viral across developer communities, Boris Cherny shared statistics that seemed almost unbelievable:
"For me personally, it has been 100% for two+ months now, I don't even make small edits by hand. I shipped 22 PRs yesterday and 27 the day before, each one 100% written by Claude." — Boris Cherny, Creator of Claude Code
To put this in perspective: most developers consider shipping 2-3 well-crafted pull requests per day a productive output. Cherny is shipping 10x that volume—without typing a single line of code himself.
The Numbers That Matter
| Metric | Boris Cherny's Output |
|---|---|
| PRs shipped daily | 22-27 |
| Code written by hand | 0% |
| Duration of this workflow | 2+ months |
| Small manual edits | None |
What "100% AI-Written" Actually Means
When Cherny says 100% AI-written, he means exactly that. But this doesn't mean he's passive in the process. His role has shifted from writing code to directing code.
The New Developer Workflow
Traditional Development:
- Understand the requirement
- Research solutions
- Write code line by line
- Debug issues
- Write tests
- Submit for review
Cherny's AI-First Development:
- Understand the requirement
- Describe the solution to Claude
- Review AI-generated code
- Direct Claude to fix issues
- Have Claude generate tests
- Submit for review
The mental work remains—understanding problems, designing solutions, reviewing quality. But the mechanical work of typing code has been completely eliminated.
How Cherny Directs Without Typing
Cherny's approach relies on several key practices:
1. Plan Mode First He starts most sessions in "Plan Mode" (accessible by pressing Shift+Tab twice in Claude Code), iterating with Claude until the approach is solid. Only then does he switch to auto-accept mode for implementation.
2. Clear Communication Instead of vague requests, he provides specific requirements, constraints, and context. The clearer the direction, the better the output.
3. Iterative Refinement If output isn't right, he doesn't manually fix it—he directs Claude to fix it. This maintains the 100% AI-written standard while ensuring quality.
Company-Wide Adoption at Anthropic
Cherny's personal workflow isn't an outlier. Across Anthropic, AI-generated code has become the norm.
"Pretty much 100%" of code across Anthropic is AI-generated, according to Cherny. An Anthropic spokesperson clarified the company-wide figure is between 70% and 90%.
Why Claude Code at Anthropic is Special
There's a particularly interesting detail: about 90% of Claude Code's own code is written by Claude Code. The tool is essentially building itself.
This creates a powerful feedback loop:
- Claude Code writes code for Claude Code
- Improvements make Claude Code better at writing code
- Better code makes Claude Code more capable
- The cycle accelerates
The "Unshackled" Feeling
"Engineers just feel unshackled, that they don't have to work on all the tedious stuff anymore." — Boris Cherny
This quote captures something important about the psychological shift. Developers who adopt AI-first workflows often report feeling liberated from the mechanical aspects of coding, allowing them to focus on what they actually enjoy: solving problems and building products.
What This Means for the Industry
Cherny made a prediction that every developer should consider:
"I think most of the industry will see similar stats in the coming months—it will take more time for some vs others. We will then start seeing similar stats for non-coding computer work also."
The Trajectory is Clear
| Timeline | Expected Industry Adoption |
|---|---|
| Now | Early adopters (70-100% AI-generated) |
| 6 months | Fast followers reach similar levels |
| 1-2 years | Majority of industry adopts AI-first |
| Beyond | Extends to non-coding work |
What This Means for Different Developers
For Individual Contributors: The ability to ship 10x more while maintaining quality creates massive career advantages for early adopters.
For Tech Leads: Team management shifts from coordinating code writers to orchestrating AI-augmented developers.
For Organizations: Smaller teams can achieve what previously required large engineering departments.
How to Start Your Own AI-First Coding Journey
Ready to adopt a similar workflow? Here's how to begin:
Step 1: Set Up Claude Code
- Install Visual Studio Code
- Add the Claude extension
- Connect your Anthropic account
- Start with simple tasks to build familiarity
Step 2: Adopt Plan Mode Discipline
Before implementing anything:
- Enter Plan Mode (Shift+Tab twice)
- Describe what you want to build
- Iterate on the plan until it's solid
- Only then switch to implementation
Step 3: Resist Manual Edits
This is the hardest part. When you see a small issue, your instinct will be to fix it manually. Instead:
- Describe the issue to Claude
- Let Claude make the fix
- Review the change
This maintains your feedback loop and helps Claude learn from your preferences.
Step 4: Build Your CLAUDE.md
Document mistakes and preferences so Claude learns:
- Style conventions
- Common patterns in your codebase
- Mistakes to avoid
- Best practices for your project
Step 5: Use Verification Loops
Give Claude ways to verify its own work:
- Run tests after changes
- Check UI in browser
- Validate against requirements
Cherny calls this "probably the most important thing" for getting great results.
The Mindset Shift Required
Adopting a 100% AI-coded workflow requires more than new tools—it requires a new mindset.
From Craftsman to Director
Traditional programming valued craftsmanship: elegant algorithms, clean code, efficient implementations. These still matter, but your role shifts from craftsman to director.
You're no longer the one wielding the tools. You're the one deciding which tools to use, what to build, and evaluating whether the result meets the standard.
From Typing to Thinking
When you stop typing code, something interesting happens: you have more time to think. About architecture. About user experience. About whether you're solving the right problem.
This is where human value increasingly lies—not in implementation, but in direction and judgment.
From Individual Output to Multiplied Impact
Cherny ships 22-27 PRs per day. That's not humanly possible with traditional coding. But it's possible when you're directing AI rather than doing everything yourself.
Your impact is no longer limited by your typing speed or time. It's limited by your ability to direct, review, and ensure quality.
Frequently Asked Questions
Does Boris Cherny really not write any code by hand?
Yes. He's been clear that for over two months, 100% of his code has been AI-generated—including small edits that most developers would do manually.
What model does he use?
Cherny uses Opus 4.5 with thinking enabled for all his coding work, valuing quality over speed.
Is this workflow only possible at Anthropic?
No. While Anthropic engineers have advantages in early access and deep familiarity with the tools, the core workflow is replicable by anyone with access to Claude Code.
How long does it take to adopt this workflow?
Most developers report the transition taking 2-4 weeks of consistent use. The key is resisting the urge to make manual edits and trusting the iterative process with AI.
What happens to coding skills?
Understanding code remains essential—you need to review, evaluate, and direct effectively. But the mechanical skill of writing code becomes less important than the strategic skill of directing AI.
Bottom Line
Boris Cherny's workflow represents the cutting edge of AI-assisted development. While not everyone will immediately achieve 100% AI-generated code, the direction is clear:
The developers who thrive will be those who learn to direct AI effectively, not those who resist the transition.
Key takeaways:
- 100% AI-coded workflows are achievable today
- The shift is from writing code to directing code
- Planning and verification are more important than ever
- Early adopters gain significant productivity advantages
- This trend will accelerate, not slow down
The question isn't whether to adopt AI-first development—it's how quickly you can make the transition.
Want to accelerate your development with AI-first practices? Contact Houston IT Developers to learn how we help teams adopt cutting-edge AI development workflows.
Sources:

Houston IT Developers
Houston IT Developers is a leading software development and digital marketing agency based in Houston, Texas. We specialize in web development, mobile apps, and digital solutions.
View all posts →Need Help With Your Project?
Our team of experts is ready to help you build your next web or mobile application. Get a free consultation today.
Get in Touch