AI Coding Tools Alter Language Choices
AI coding assistants are creating "convenience loops" that change which programming languages developers use, according to a new analysis of GitHub data. Engineers are doubling down on familiar languages where AI can quickly generate boilerplate, even if other languages might be more performant. This suggests "prompt fluency" is becoming a key skill alongside traditional coding ability.
The rise of AI coding assistants has directly fueled a major shift in programming language popularity, with TypeScript overtaking Python to become the most-used language on GitHub for the first time in August 2025. This surge is attributed to TypeScript's strong type system, which provides clear guardrails for AI tools, helping to catch errors in generated code before it reaches production. This phenomenon is what GitHub Developer Advocate Andrea Griffiths calls a "convenience loop": AI tools reduce the friction of using a particular technology, which in turn drives its adoption. As more developers use TypeScript, they generate more training data for AI models, making the AI even better at writing TypeScript, thus reinforcing the loop. This has led to a massive 66% year-over-year growth in TypeScript contributors on GitHub. While TypeScript's growth is significant in general development, Python remains the undisputed leader for AI and data science-specific projects. In 2025, nearly half of all new AI repositories on GitHub were initiated with Python, which saw its number of contributors grow by 48% to 2.6 million. The language's extensive ecosystem of libraries like TensorFlow and PyTorch continues to make it the default for model training and exploration. This new landscape elevates "prompt fluency"—the ability to write effective instructions for AI models—to a core developer skill, rivaling traditional coding ability. The quality of AI-generated code is directly dependent on the precision of the prompt, demanding that engineers specify context, patterns, and constraints to receive usable, production-grade output. The trend benefits established languages with vast amounts of public code for training AI models. According to TypeScript's lead architect, Anders Hejlsberg, new languages are at a disadvantage because AI's ability to write code is proportional to the volume of examples it has seen. This creates a potential "frozen in time" effect, where dominant languages become further entrenched. Despite the productivity gains, with some teams coding up to 40% faster, trust in AI-generated code remains low. The 2025 Stack Overflow Developer Survey found that while 84% of developers use or plan to use AI tools, only 29% trust the accuracy of the output, and 45% report losing significant time debugging AI-generated code. This dynamic is even changing how language popularity is measured. With developers increasingly using private conversations with LLMs instead of public forums, traditional metrics like Stack Exchange questions are dwindling. In 2025, weekly questions on Stack Exchange for top programming languages dropped to just 22% of 2024 levels.
