The Past, Present and Future of Programming Languages by Kevlin Henney - Voxxed Days CERN '26
(link)Summary
Kevlin Henney traces the evolution of programming languages from early compiler and language-design milestones to today’s dominant ecosystems and the emerging influence of AI. Using historical references such as Fortran, Lisp, ALGOL, and the von Neumann model, he argues that modern languages mostly evolve by borrowing features from earlier ones rather than being replaced outright. The talk compares language popularity across multiple rankings, including Tiobe, RedMonk, and the IEEE Spectrum list, to show how stable the top mainstream languages remain over time. It also discusses the biases in those rankings, the shrinking role of Stack Overflow as a signal, and why older languages continue to dominate because of installed base, tooling, and ecosystem gravity. A final theme is the effect of large language models on programming: they favor languages with abundant training data such as Python, JavaScript, and TypeScript, and increasingly rely on generated code for tasks they handle poorly. Henney closes by arguing that while new languages are easier than ever to create, the mainstream set will stay surprisingly stable.
Key Takeaways
- The dominant programming languages change slowly; the top mainstream set has remained remarkably stable across multiple rankings.
- Historical language design ideas from Fortran, ALGOL, Lisp, and functional programming still shape modern language features.
- Popular rankings such as Tiobe, RedMonk, and IEEE Spectrum each have different biases and should be interpreted carefully.
- Stack Overflow traffic is no longer a reliable proxy for developer language popularity on its own.
- AI systems tend to favor languages with large amounts of training data and often generate code to compensate for weaknesses in arithmetic, parsing, or counting tasks.
- Creating a new language is easier than ever, but ecosystem support and installed code bases make mainstream disruption unlikely.
Sections
Historical context and language design
Henney starts with early language history, including Fortran as an ambitious compiler experiment and ALGOL as a source of many ideas now taken for granted. He highlights how features such as block structure, expression-oriented programming, `if-then-else`, and `for` loops emerged gradually and were often invented in response to implementation and notation problems. He also revisits John Backus’s critique of von Neumann-style programming and the push toward more mathematical and declarative approaches.
How programming language popularity is measured
The talk compares Tiobe, RedMonk, and IEEE Spectrum rankings to show that the same small set of languages repeatedly occupies the top positions. Henney emphasizes that each ranking has its own bias: search-engine visibility for Tiobe, Stack Overflow and GitHub activity for RedMonk, and broader survey-based signals for IEEE Spectrum. He shows that the top five languages have been especially stable for years, and that the distribution is heavily skewed rather than linear.
Why the mainstream set stays stable
Henney argues that the main reason language change is slow is the accumulated weight of existing code, tooling, and developer habit. Languages tend to borrow features from one another rather than being replaced. He frames modern language evolution as a process of feature adoption—lambdas, comprehensions, streams, pattern matching, and expression-oriented constructs—often copied from functional or logic programming traditions while the language name and ecosystem remain the same.
Functional ideas in mainstream languages
A major theme is that many modern languages have absorbed functional concepts even when they are not purely functional. Henney points to lambdas, closures, comprehensions, and expression-oriented style as examples of ideas borrowed from functional programming. He also notes that SQL is declarative and logic-based, and that Excel’s formula language is effectively the most widely used functional language, especially after Microsoft added `LAMBDA` support.
AI, code generation, and the future
Henney closes by discussing how large language models change the programming-language landscape. AI systems are strongest in languages with abundant training data such as Python, JavaScript, and TypeScript, while newer or less common languages are disadvantaged. He also notes that LLM-based systems increasingly rely on generated code or tool orchestration for tasks like arithmetic, string counting, and CSV handling. His conclusion is that AI will amplify existing trends rather than overturn the mainstream set of languages.
Keywords: programming language evolution, kevlin henney, voxxed days cern, tiobe index, redmonk ranking, ieee spectrum top programming languages, fortran history, algol 68, lisp, functional programming, lambda calculus, von neumann architecture, stack overflow trends, github language popularity, excel lambda, python, javascript, typescript, ai code generation, large language models, language design, compiler history, programming paradigms