Key Takeaways
- The 100% Goal: Anthropic ’s CEO predicted this a year ago; today, it is a reality for their engineering teams.
- Human Roles Shift: Employees aren’t “coders” anymore; they are “reviewers” and “directors” who check the AI’s work.
- Speed Explosion: Engineers are now finishing in days what used to take months, sometimes sending 27 major updates in a single day.
- Market Panic: Investors are worried that traditional software companies will become obsolete, leading to a $285 billion loss in stock values this week.
Anthropic Reaches 100% AI-Written Code
The future of making software just changed forever: Anthropic ’s top leaders have confirmed that their internal systems are now almost entirely built by their own AI.
- What Happened: Mike Krieger, Anthropic ’s Chief Product Officer, announced that “effectively 100%” of their internal code is now written by their AI, Claude.
- Why It Matters: This marks the first time a major tech giant has reached a point where humans have stopped writing the actual lines of computer instructions.
- What’s Next: Expect a massive shift in the job market as other companies try to copy this “hands-off” approach to building technology.
This article explores how we reached this milestone, what it looks like inside the world’s most advanced AI lab, and why the “SaaSpocalypse” (the collapse of traditional software services) is causing panic in the stock market.
Claude Writing Claude: The Milestone Explained
On February 9, 2026, during the Cisco AI Summit, Mike Krieger shared a “vignette” from inside Anthropic. He revealed that their internal engineering has hit a milestone once thought impossible: Claude is now writing Claude. Just one year ago, Anthropic’s CEO, Dario Amodei, predicted that 90% of code would be AI-generated within a year. Many skeptics called it “crazy.” However, Krieger confirmed that for most products at Anthropic, that number has surpassed the prediction and is now effectively 100%.
”Claude products and Claude code are being entirely written by Claude,” Krieger stated during his talk.
This means the engineers at Anthropic have stopped typing out the complex “languages” computers speak. Instead, they describe what they want to the AI, and the AI builds the entire system itself.
How the New Way of Working Works
At Anthropic, the day-to-day life of an engineer has transformed. Instead of staring at a screen and typing for eight hours, they act more like a movie director or a head chef.
| Role | Old Way (2024) | New Way (2026) |
| Primary Task | Writing lines of code by hand | Describing goals and giving “prompts” |
| Daily Output | 1–2 small updates | 20–27 major updates (pull requests) |
| Main Tool | Keyboard and coding software | AI Agents (Claude Code) |
| Human Value | Knowledge of computer languages | Architectural vision and safety checks |
A lead engineer at the company, Boris Cherny, recently shared that he hasn’t made even a “small edit” by hand in over two months. He reported shipping 27 major code contributions in a single day, a feat that would normally take a human team weeks of effort.
The “SaaSpocalypse” and the $285 Billion Crash
While this is a technical victory for Anthropic, it has sent a shockwave of fear through the global economy. If a handful of people can use AI to build world-class software in days, what happens to the massive companies that employ thousands of people to do the same thing slowly?
This week, the financial world coined a new term: the “SaaSpocalypse.” As news of Anthropic’s efficiency went viral, investors began dumping shares of traditional software-as-a-service (SaaS) companies.
- Total Market Loss: Roughly $285 billion was wiped off the value of software stocks in a single day.
- Affected Giants: Major IT service companies in India (like TCS and Infosys) saw their stocks fall as investors worried that AI “agents” would replace their human workforces.
- The Logic: If the AI can now directly perform tasks like sales, legal filing, and data analysis tasks that used to require expensive software subscriptions, those old software companies may soon have no customers left.
What the Future Holds: From “Writer” to “Editor”
The shift at Anthropic shows that the “bottleneck” in technology is no longer the act of writing. The bottleneck is now judgment.
As Mike Krieger noted at the summit, the biggest risk is no longer a lack of talent, but “human processes getting in the way.” Companies that refuse to let AI handle the “typing” will likely be replaced by smaller, faster teams that use AI to do 100% of the heavy lifting.
For the average worker, this means learning how to “orchestrate” AI. The most valuable skill in 2026 isn’t knowing how to code; it’s knowing how to verify that the AI’s work is safe, accurate, and aligned with what the business needs.
FAQ: Everything You Need to Know
1. Does 100% AI code mean humans are being fired?
Not necessarily. At Anthropic, engineers are still there, but they have moved from “manual labor” to “high-level oversight.” They spend their time thinking about the “big picture” rather than fixing typos in code.
2. Is AI-written code safe?
Anthropic uses “adversarial” AI to check its own work. One AI writes the code, and another AI tries to find bugs or safety issues in it. This “self-checking” loop is often more accurate than human review.
3. Will this happen at every company?
Anthropic is an AI company, so they are the first to adopt this. However, with the release of their newest tool, Claude Opus 4.6, they are making these “auto-coding” abilities available to everyone.
4. What is the “SaaSpocalypse”?
It refers to the potential collapse of traditional software companies whose services can now be performed more cheaply and quickly by independent AI agents.
