You Think Languages Change Fast. Here’s the Data That Proves Otherwise.

Henney makes a data-backed argument that mainstream programming language adoption is far slower than the industry narrative suggests. Using three independent rankings (TIOBE, RedMonk, IEEE Spectrum), he shows the top 10 is dominated by languages from the 20th century. Python, Java, C++, and JavaScript have held the top spots for years. The tail is long,

but the head is remarkably stable. Most new languages are out in the long tail. Even Go and Rust, often considered new, are over a decade old. The core innovations—control structures, coroutines, functional features—date to the 1960s and 70s. Languages today borrow heavily from old ideas rather than inventing new ones. The rate of change

in the mainstream has slowed because there is too much existing code and too many jobs built around it. The major twist: AI LLMs are now consumers of programming languages. They write code in proportion to the volume of training data they've seen. This creates a feedback loop that entrenches Python, JavaScript, and TypeScript. Anders

Hejlsberg is cited: new languages are systemically disadvantaged because AI can't generate reliable code for them without massive datasets. The practical implication is that teams should not bet on language disruption. The barrier to entry for a new language in the mainstream is higher than ever. Investment in tooling, training, and AI integration for existing