Are You Still Using Python? – OSOS/Omega Blog

“In August 2025, TypeScript overtook both Python and JavaScript.”
– GitHub, Octoverse 2025¹

According to GitHub’s Octoverse 2025 report, TypeScript became the most-used language on GitHub in August 2025, measured by the number of monthly contributors. GitHub describes this as the most significant shift in language rankings in over a decade, with AI playing a key role.

I build my own AI agents in TypeScript. That’s why I’m surprised by how rarely I see other engineers do the same. Almost everyone I talk to still reaches for Python by default.

Don’t get me wrong: Python is a great language, and there are good reasons for its dominance in AI. But I think we’re too quick to treat it as the default choice for building AI agents.

So before you start your next AI project, take a moment to ask yourself: Are you choosing Python because it’s the best tool for the job – or simply because it’s what you already know?

Still pushing everything through the Python tunnel?

Where the Preference for Python Comes From

There’s a good reason Python became the default choice for AI. Before the current AI boom, there was the data science boom. Python emerged as the go-to language for data analysis, and when developers started experimenting with training their own neural networks, it was hard to avoid. Its ecosystem of libraries and tools made it the natural choice.

That legacy continues today. Python still dominates AI and data science. According to GitHub, nearly half of all new AI repositories are written in Python.¹

But I think there’s another, more fundamental reason Python remains the default for AI agents – and arguably the more important one: most engineers building AI agents don’t come from a web development background.

For them, TypeScript is unfamiliar territory. Python is what they know, so Python is what they use. And that makes sense – until you start looking at what building AI agents actually involves.

Python Had a Head Start

Python didn’t become the default language for AI agents by accident. Some of the most popular frameworks launched in Python first, giving it a head start that reinforced its position. LangChain, for example, was released as a Python library in October 2022, with the TypeScript version following four months later. OpenAI’s Agents SDK followed a similar path: the Python version launched in March 2025, three months before its TypeScript counterpart.²

But that early lead no longer tells the whole story.

Today, the major building blocks for AI agents are available in TypeScript, too. LangChain and LangGraph both support it, as do OpenAI’s Agents SDK and Anthropic’s Claude Agent SDK. The Model Context Protocol (MCP), which allows agents to interact with tools and external data sources, has an official TypeScript SDK. And Vercel’s AI SDK provides a framework specifically designed for building AI-powered applications on the web.³

In other words, you no longer need Python to build serious AI agents. The ecosystem is there. The question is whether more developers will start using it.

What Is TypeScript, Anyway?

If you’re not familiar with web development, TypeScript might sound like just another programming language to learn. In reality, it’s a natural extension of something you may already know: JavaScript, the language that powers the web.

Developed by Microsoft and first released in 2012, TypeScript adds static types to JavaScript. Instead of letting a variable hold just about anything, you can specify what kind of value it should contain – a number, a string, or a particular object with defined properties. The compiler can then catch many mistakes before the code runs.

TypeScript code is ultimately compiled into plain JavaScript, so it runs wherever JavaScript does: in the browser, on a server, or in serverless and edge environments. You get the benefits of a more structured language without giving up the JavaScript ecosystem.

And that matters for AI agents. According to GitHub, stronger typing is one factor behind TypeScript’s growing popularity: types help developers catch errors earlier, including mistakes in code generated by AI models. When an AI agent writes code or works with complex data structures, having the compiler flag problems before runtime can make a real difference.

If Your Agent Has a Browser UI, TypeScript Is Already Part of the Picture

If you’re building a small agent for personal use, Python is perfectly fine. You don’t need to introduce a new language just for the sake of it.

But what if you want to give your agent a proper browser interface? Or make it available to your team? Or integrate it into an existing web application?

At that point, TypeScript becomes hard to avoid. And if you’re using it for the frontend anyway, why maintain your backend in a different language?

Using TypeScript across the stack means your frontend and backend share the same type definitions. If the structure of an API response changes, the compiler can tell you exactly where the frontend needs to be updated. No hunting for mismatched fields at runtime, no maintaining separate definitions of the same data structures.

For applications that combine AI agents with web interfaces, that can make development considerably simpler.

“But I’d Have to Learn a New Programming Language”

I don’t think that’s a convincing argument anymore. In fact, I’d go a step further: with AI coding agents, you no longer need to master a programming language to build useful software.

I understand the hesitation. Watching an AI agent like Claude Code write hundreds of lines of code in a language you barely know can feel uncomfortable. How can you judge whether the code is any good?

The answer is to focus less on syntax and more on how software works. For building useful tools, understanding the architecture and the security implications is often more important than memorizing the rules of a particular language.

For example:

Frontend vs. backend: What runs in the browser, and what must run on the server?
Authentication and authorization: Who can sign in, and who is allowed to access which data?
Data and storage: Where is your data stored, and how does your application access it?
Secrets and credentials: Which API keys and other credentials must stay on the server and never be exposed to the browser?

These are the questions that matter when you’re building software with AI. If you understand them, you can give your coding agent better instructions, make informed architectural decisions, and spot dangerous mistakes.

Imagine an agent putting an API key directly into frontend code, where every visitor could inspect it. You don’t need to be a TypeScript expert to recognize the problem. You need to understand how web applications work.

Let AI handle the syntax. You focus on the architecture, the security, and the problem you’re trying to solve.

Structure of a web application with an AI agent: TypeScript runs from the browser to the database, Python only on the server.

So, Which Language Should You Choose?

Python is still a great choice, especially for data science, machine learning, and quick experiments. But if you’re building AI agents that will grow into web applications, TypeScript deserves a serious look.

Don’t choose Python just because it’s the traditional language for AI. And don’t choose TypeScript just because it’s trending on GitHub. Choose the language that fits the product you’re building – and the ecosystem you want to build it in.

Where Do You Stand?

Curious about where you stand with your AI skills? Take our 5-minute quiz to find out.

In our live online trainings at the OSOS Academy, we meet you where you are and help you take the next step toward building your own AI agents.

And now I’m curious about you: Which language do you use to build your AI agents – Python, TypeScript, or something else? And why?

Sources

GitHub: “Octoverse: A new developer joins GitHub every second as AI leads TypeScript to #1.” GitHub Blog, October 2025. The count is based on monthly active contributors; rankings using other methodologies still place Python or JavaScript ahead. ↩ ↩²
Release histories: LangChain on PyPI (Python, first version October 2022) and npm (TypeScript, February 2023), see also the LangChain blog post “TypeScript Support” from 17 February 2023; OpenAI Agents SDK on PyPI (March 2025) and npm (June 2025). ↩
TypeScript versions on GitHub: LangChain.js, LangGraph.js, OpenAI Agents SDK, Claude Agent SDK, MCP TypeScript SDK; plus Vercel’s AI SDK. ↩