AI writes up to 30% of the Microsoft's code, CEO says

Microsoft CEO Satya Nadella announced that AI now generates up to 30% of the company’s codebase, with the proportion increasing over time.

He shared this during a live discussion with Meta CEO Mark Zuckerberg at LlamaCon, Meta’s AI developer event in Menlo Park, California.

Zuckerberg revealed Meta’s plans to create an AI model capable of handling a significant portion of the company’s future AI systems.

He estimates AI could manage half of Meta’s software development within the next year.

This reflects a broader trend, with companies like Google, Shopify, and Duolingo also exploring AI to reduce human involvement in software development.

🔗 Source: CNBC

🧠 Food for thought

1️⃣ The evolution of programming productivity follows a consistent pattern

Microsoft’s 30% AI-written code represents the latest step in a decades-long progression toward higher levels of programming abstraction.

In the 1950s and 60s, programmers wrote assembly code, where a single line might represent just one CPU instruction; by the 1970s, higher-level languages like C allowed programmers to express algorithms more concisely, increasing productivity dramatically1.

The 1990s and 2000s saw open source libraries and frameworks that enabled developers to reuse entire code components rather than writing them from scratch, with studies showing productivity gains of 40-60% through code reuse1.

Today’s AI coding assistants like GitHub Copilot (Microsoft) and AlphaCode (Google) represent a similar leap, where programmers can express intent in natural language and have AI generate working code. This continues the historical trend of abstracting away implementation details.

This progression suggests that rather than replacing programmers, AI tools are enabling developers to shift focus to higher-level problems while automation handles increasingly complex implementation tasks.

2️⃣ A new competitive frontier emerges in AI coding tools

The major tech companies are racing to establish dominant positions in AI-assisted programming, recognizing it as a strategic technology.

Microsoft leads with both significant internal adoption (20-30% AI-written code) and external products like GitHub Copilot, which reported over 1.3 million paying subscribers in 20231.