Anthropic's Claude Writes 80% of Internal Code, Boosts Optimization Speed

@nathanbenaich· July 24, 2026 View original

Summary

Anthropic reports that its AI model, Claude, now writes over 80% of the company's merged code, significantly accelerating development. Kernel optimization speedups have dramatically increased from 3x to 52x within a year, indicating a sustained, rapid improvement curve.

Anthropic, a leading AI research company, has revealed impressive internal adoption of its own AI model, Claude, for software development. According to Ted Moskovitz, over 80% of the code merged into Anthropic's codebase is now generated by Claude, showcasing a profound integration of AI into their engineering workflow. This high level of AI-assisted coding is not only streamlining development but also yielding substantial performance gains. The company has observed an extraordinary acceleration in kernel optimization speedups, jumping from approximately 3x to an astounding 52x in just one year. Moskovitz emphasized that this rapid improvement curve shows no signs of slowing down, suggesting a continuous upward trajectory in AI's ability to enhance efficiency and performance in complex technical tasks. He also highlighted that the true measure of AI acceleration goes beyond simple benchmarks, and that larger models often prove to be more cost-effective in the long run.

Why it matters

This demonstrates the tangible productivity gains and efficiency improvements that advanced AI models can bring to software development and complex engineering tasks, setting a new benchmark for AI integration.

How to implement this in your domain

  1. 1Pilot AI code generation tools for specific modules or non-critical projects.
  2. 2Invest in training engineers to effectively prompt and review AI-generated code.
  3. 3Establish metrics to track productivity and quality improvements from AI coding assistance.
  4. 4Explore AI-driven optimization tools for performance-critical code sections.
  5. 5Develop internal guidelines for responsible AI code integration and security.

Who benefits

Software DevelopmentTechAI/MLCloud Computing

Key takeaways

  • AI models can now generate a significant majority of production-level code.
  • AI is dramatically accelerating complex optimization tasks, like kernel tuning.
  • The benefits of AI in engineering are scaling rapidly and show no signs of slowing.
  • Larger AI models can be more cost-effective due to their efficiency gains.

Original post by @nathanbenaich

"The science of scaling with @ted_moskovitz of @anthropic at @raais 2026: Over 80% of Anthropic's merged code is now written by Claude. Kernel-optimization speedups went from ~3x to ~52x in a single year, and "we have not seen that curve bend." Ted on the real measure of AI accele…"

View on X

Originally posted by @nathanbenaich on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses