For decades, the choice of programming language for a new software project was governed by a relatively stable set of factors: performance requirements, team expertise, library availability, and the legacy systems a project needed to interface with. But a new variable has entered the equation — one that is subtly but measurably altering the calculus for engineering teams worldwide. The rise of AI-powered coding assistants, from GitHub Copilot to Anthropic’s Claude and OpenAI’s ChatGPT, is beginning to influence which programming languages developers choose for new projects, and the implications for the software industry could be profound.
The question was raised prominently in a recent discussion on Slashdot, where developers and technologists debated whether AI tools are meaningfully shifting language adoption patterns. The thread drew hundreds of comments from working programmers, many of whom reported firsthand experience with the phenomenon. The consensus, while not unanimous, pointed in a clear direction: AI code generation works better in some languages than others, and that asymmetry is starting to matter.
Python’s Dominance Gets an AI Tailwind
Python, already the world’s most popular programming language by several measures, appears to be the biggest beneficiary of the AI coding era. The language’s extensive presence in training data — owing to its use in education, data science, web development, and machine learning itself — means that large language models produce notably higher-quality Python code than they do for less common languages. Multiple developers in the Slashdot discussion noted that AI assistants generate Python suggestions that are more idiomatic, more correct, and require less manual editing than output in languages like Rust, Haskell, or even C++.
This creates a self-reinforcing cycle. As AI tools get better at Python, more developers choose Python for new projects, which generates more Python code for future model training, which further improves AI performance in Python. The dynamic is reminiscent of network effects in platform economics, where early advantages compound over time. For languages with smaller codebases in public repositories, the opposite risk looms: AI tools perform worse, developers gravitate away, and the training data gap widens.
The Training Data Problem for Niche Languages
The quality of AI-generated code is directly tied to the volume and quality of training data available for a given language. Languages like COBOL, Fortran, Ada, and even relatively modern but less mainstream languages like Elixir or OCaml have far less representation in the open-source repositories that form the backbone of LLM training sets. Developers working in these languages report that AI assistants frequently produce incorrect or subtly broken code, sometimes hallucinating APIs that don’t exist or generating patterns that violate the language’s conventions.
This disparity has real business consequences. A commenter on the Slashdot thread, identifying as a senior engineer at a financial services firm, described a recent architecture decision where the team chose TypeScript over a more specialized language partly because “the AI tools just work better with it.” The calculus was straightforward: if AI assistants can handle boilerplate and routine coding tasks more effectively in TypeScript, the team’s overall productivity would be higher, even if the specialized language might have been technically superior for the use case.
JavaScript and TypeScript Hold Strong
JavaScript and its typed superset TypeScript occupy a privileged position similar to Python’s. The sheer volume of JavaScript code on GitHub and Stack Overflow ensures that AI models have been trained on an enormous corpus of examples spanning every conceivable use case. Developers report that tools like GitHub Copilot are particularly effective at generating React components, Express.js server code, and other common JavaScript patterns. TypeScript benefits additionally because its type annotations give AI models more context to work with, resulting in suggestions that are more likely to be correct.
The TIOBE Index, which tracks programming language popularity, has shown Python and JavaScript maintaining their top positions throughout 2024 and into 2025. While multiple factors drive these rankings, industry analysts have begun to acknowledge AI tooling as a contributing variable. According to analysis shared by TIOBE, languages that are well-supported by AI development tools tend to see stronger adoption trends, particularly among newer developers who have grown up using these assistants as a default part of their workflow.
Rust, Go, and the Middle Ground
Not all newer languages are disadvantaged by the AI shift. Rust and Go, both of which have seen significant growth in recent years, occupy an interesting middle ground. Go’s relatively simple syntax and limited feature set make it a language that AI models can generate reasonably well, even with a smaller training corpus than Python or JavaScript. The language’s design philosophy — favoring explicitness and simplicity over abstraction — aligns well with the pattern-matching strengths of current LLMs.
Rust presents a more complex picture. The language’s strict ownership and borrowing rules mean that AI-generated Rust code frequently fails to compile, even when the logic is conceptually correct. Several developers in the Slashdot discussion noted that while Copilot and similar tools can produce useful Rust snippets, the error rate is noticeably higher than for Python or Go. However, some argued this is less problematic than it sounds: Rust’s compiler is famously helpful at identifying and explaining errors, so the combination of AI-generated code plus compiler feedback can still be productive. Others pointed out that as Rust’s popularity grows and more Rust code enters training datasets, AI performance in the language should improve.
Enterprise Decision-Making Feels the Pull
The influence of AI on language choice is perhaps most consequential at the enterprise level, where technology decisions affect hundreds or thousands of developers and persist for years. Chief technology officers and engineering directors are increasingly factoring AI tooling support into their technology stack evaluations. A language that enables a 20% productivity boost through better AI assistance represents a significant competitive advantage when multiplied across a large engineering organization.
Stack Overflow’s annual developer survey has tracked growing AI tool adoption among professional developers. According to Stack Overflow’s 2024 survey, over 76% of developers reported using or planning to use AI tools in their development process. Among those developers, satisfaction with AI-generated code varied significantly by language, with Python, JavaScript, and TypeScript users reporting the highest satisfaction rates. This data point, while not establishing direct causation, aligns with the pattern described by developers in forums and industry discussions.
The Risk of a Homogenized Future
Critics of this trend warn about the potential for AI to create a monoculture in programming languages. If developers increasingly cluster around the languages that AI tools handle best, the industry could lose the diversity of thought and approach that different programming paradigms provide. Functional programming languages like Haskell and Clojure, logic programming languages like Prolog, and systems programming languages like Zig each embody different ways of thinking about computation. If these languages lose mindshare because AI tools don’t support them well, the argument goes, the industry loses valuable intellectual tools.
There is also a quality concern. AI models trained predominantly on Python and JavaScript code absorb the patterns and anti-patterns of those communities. If developers increasingly rely on AI-generated code in these languages, they may inadvertently propagate common mistakes and suboptimal patterns at scale. The homogeneity of AI-generated code — its tendency toward the most statistically common solution rather than the most elegant or efficient one — could lead to a gradual flattening of code quality across the industry.
What Comes Next for Language Designers and Tool Builders
Language designers are not passive observers of this trend. Several programming language communities have begun actively working to improve their representation in AI training data. The Rust Foundation, for example, has invested in documentation and example code that could improve AI model performance. Similarly, some language communities are exploring ways to create curated, high-quality training datasets that would help AI models generate better code in their languages.
On the tool side, companies building AI coding assistants are aware of the language disparity and are working to address it. Anthropic, OpenAI, and Google have all made efforts to improve their models’ performance across a wider range of languages. Specialized fine-tuning on underrepresented languages is one approach; another is incorporating language-specific tooling, such as compilers and type checkers, into the AI generation pipeline to catch errors before they reach the developer.
The interplay between AI capabilities and programming language choice represents one of the most significant second-order effects of the generative AI wave in software development. While the technology press has focused heavily on whether AI will replace programmers — a question most working developers regard as premature — the more immediate and measurable impact may be on which languages those programmers write in. For enterprise technology leaders, language community maintainers, and individual developers planning their career trajectories, this shift deserves close attention. The programming languages that thrive in the next decade may be determined not just by their technical merits, but by how well they play with the AI tools that are rapidly becoming standard equipment in every developer’s toolkit.