Python’s long game: Guido van Rossum on growth, AI, and staying readable

When GitHub’s 2025 Octoverse data landed, it showed TypeScript overtaking Python as the most-used language on the platform—the biggest shift in more than a decade. But the same report showed Python still growing 49% year over year, holding its ground as the default choice in AI, science, and education. The creator of Python, Guido van Rossum, was genuinely caught off guard by the number, noting how it diverges from other trackers like the TIOBE Index.

That growth, and the language’s continued dominance in key fields, traces back to Python’s origins as a pragmatic response to the limits of existing tools.

A reaction to C’s friction

Python emerged from Van Rossum’s work on a novel operating system where C was the only language available. He found that even basic utilities—like reading two lines of input—became exercises in buffer overflow management and manual memory allocation. Shell scripts lacked the expressiveness he needed; C was too brittle for the workflow he wanted.

I wanted something that was much safer than C, and that took care of memory allocation, and of all the out of bounds indexing stuff, but was still an actual programming language.

The result was a language designed to handle those hard parts so developers could focus on the problem at hand. From the start, Python’s DNA included clarity, friendliness, and minimal friction—values that now underpin its role in AI, data science, and enterprise development.

The name itself was a deliberate choice. Van Rossum wanted “a little irreverence” against the solemnity of computer languages, picking Monty Python’s Flying Circus as the namesake. That sense of accessibility, paired with readable syntax and welcoming error messages, has helped flatten the learning curve for countless newcomers.

Why Python became the language of AI

Python’s hold on AI isn’t a happy accident. Van Rossum points to ecosystem gravity: once a language proves useful, new software builds on it to leverage what already exists. Key projects have compound that effect over time:

  • NumPy: foundational numerical arrays
  • pandas: streamlined data manipulation
  • PyTorch: machine learning at scale
  • Local model runners and LLM agents: today’s frontier with projects like ollama leading the way

“The people now writing things for AI are familiar with Python because they started out in machine learning,” Van Rossum said. That pipeline—from scientific computing to modern AI infrastructure—has made Python more than a language for AI; it helped AI become what it is today.

No panic on typing in the LLM era

With AI generating increasing amounts of Python code, a natural question is whether the language needs stricter typing. Van Rossum’s answer is a firm no. He called Python’s optional typing system “plenty,” adding that the real issue lies in training data—most tutorials don’t teach static typing, so models don’t see enough annotated code.

AI should adapt to us, not the other way around.

He does see room for improvement on the model side: “If I ask an AI to add a type annotation, it usually researches it and gets it right.” But the philosophy holds: Python prioritizes the developer experience, and AI tools should meet developers where they are.

Built for beginners, driven by real-world need

Python’s reputation as an entry-point language stems from its forgiving nature. “There aren’t that many things you can do wrong that produce core dumps or incorrect magical results,” Van Rossum noted. Python explains what went wrong and where. He regularly hears from users that Python made their career possible—often people without formal computer science training.

“A lot of Python users and contributors do not have a computer science education … because their day jobs require skills that go beyond spreadsheets.” That pattern is visible in this year’s Octoverse data, which showed India alone adding more than 5 million developers in 2025, many from non-traditional education paths.

Every design decision has trade-offs, though. Python’s indentation-based grouping remains beloved by most, but a small group of users still emails Van Rossum directly with complaints. “Everyone else thinks that’s Python’s best feature,” he said.

Backward compatibility as a feature

Python’s longevity comes from careful evolution rather than dramatic change. The core team weighs every new feature against the risk of breaking decades of existing code. “For every new feature, we have to very carefully consider: is this breaking existing code?” Van Rossum said.

Soft keywords are a recent architectural answer to that tension—context-sensitive syntax that lets the team introduce new constructs without disrupting old programs. It’s a subtle engineering choice that keeps enterprises stable while allowing the language to keep moving forward.

Python’s trajectory offers a clear path for maintainers: solve real problems, invite input, and iterate carefully. The values that built it—readability, approachability, stability, and a touch of irreverence—remain its foundation. With AI tools now part of the workflow for most new developers on GitHub, Python’s clarity has become an even more valuable asset for turning ideas into implementations.