Something interesting is happening with Python. It's winning, and not for the reasons most people think.
The standard criticism of Python is that it's slow and badly typed. This is true. But it misses the point in the same way that early criticism of JavaScript missed the point. JavaScript was a mess compared to Java, but it didn't matter. What mattered was that it was everywhere.
Python is winning because it follows what I call the Law of Least Resistance. When given a choice between two tools, developers will nearly always choose the one that lets them ship faster. We saw this with Ruby on Rails in the 2010s. Now we're seeing it with Python, but at a much larger scale.
The most telling evidence isn't in the Stack Overflow surveys or GitHub stats, though they're impressive. It's in what the smartest founders are building. Look at the recent startups that have gained traction:
- Modal ($16M raised) turned infrastructure into Python
- Reflex ($5M raised) turned frontend into Python
- Transformers ($100M+ raised) turned complex ML into Python
- FastAPI (funded by Sequoia) turned web servers into Python
- Mojo ($130M+ raised) turned systems engineering into Python
- Streamlit (acquired for $800M) turned app development into Python
Notice the pattern? They're all taking something hard and making it easy by expressing it in Python. This isn't a coincidence.
The VPs of Engineering at big companies hate this trend. They want everything in Java or TypeScript. They have their reasons: type safety, performance, maintainability. But they're fighting a losing battle, just like the companies that tried to standardize on Angular 1.0.
The real force multiplier is Python's open source ecosystem. A handful of engineers at a startup can leverage the work of thousands of open source contributors. It's like having a huge, invisible engineering team. This is particularly evident in AI, where Python's dominance is almost total.
Many companies face this decision: they could force everything into a "serious" language like Java or Go. Instead, the successful ones often build adapter layers between Python and their main stack. It's not always the cleanest solution architecturally, but it's usually the right one pragmatically.
The most common counterargument is that Python won't scale. But this misunderstands how software works in practice. By the time you need to scale, you'll have the resources to rewrite the bottlenecks. And increasingly, those bottlenecks are being solved by projects like Mojo anyway.
What the critics miss is that Python's "weaknesses" are actually strengths in disguise. Its simplicity means more people can write it. Its dynamism means faster iteration. Its huge ecosystem means most problems are already solved.
When it comes to AI development, companies typically choose one of three paths: they either adopt Python internally, force AI engineers to use their existing stack, or build adapter layers between Python and their main codebase. The ones choosing the second option are making a bet against the entire Python ecosystem. That's a bold move, considering Python's overwhelming advantage in AI development.
The best way to predict the future is to look at what the smartest people are choosing when they're free to choose anything. Right now, that's increasingly Python. You can fight this trend, but you probably shouldn't. The winners in the AI era will be the ones who embrace it.