In my last post, I ended with a clue I hadn't fully explained yet something hiding inside Python itself that almost nobody was talking about. Here it is, for over 30 years, Python had a secret limit. Imagine buying a car with 8 powerful engines, but a rule that only lets you use one at a time forever. That's exactly what happened to one of the most popular programming languages in the world. Then, quietly, in 2025, the rule finally changed.
Part 1: Picking Up Where I Left Off
Last time, I wrote about Mojo a language that tried to fix Python's biggest weaknesses by building something completely new outside of it and ended up giving up on being Python altogether.
I said there was one more piece of the puzzle. Here it is, while Mojo was busy trying to replace Python from the outside, Python was quietly fixing itself from the inside and the story of how is honestly wilder than I expected.
Part 2: The One-Lane Highway
Modern computers are powerful. Most have 4, 8, even 16 separate "cores" think of each core as its own lane on a highway, all capable of doing work at the same time.
But for over 30 years, Python had a strange rule buried deep inside it: no matter how many lanes your computer had, Python was only allowed to drive in one of them. One instruction at a time. Ever.
This rule has a name: GIL — the Global Interpreter Lock. It's been there since Python was created in 1991.
Part 3: Why Nobody Fixed It
You'd think if something's been broken for 30 years, someone would just fix it. But the GIL wasn't a small bug it was baked into how Python worked from the very beginning.
Removing it meant risking that millions of existing Python programs, written over decades, might suddenly break. For a long time, that risk felt bigger than the reward. So, the rule just stayed. Developers learned to work around it instead of removing it.
Part 4: Someone Finally Did It Anyway
In 2023, an engineer named Sam Gross, working at Meta, proposed something nobody had managed to pull off before: a version of Python that could finally use every lane on the highway no GIL, real multi-core power, for the first time ever.
It didn't happen overnight. It rolled out carefully: first as an experimental option in Python 3.13, then officially supported in Python 3.14, released in October 2025.
Part 5: What Actually Changed
Here's the honest trade-off, because nothing is free:
->Regular single-threaded code got a little slower around 5 to 10%.
->But code that needs multiple cores at once the kind AI and data-heavy applications rely on constantly became up to 4 times faster.
That's the difference between a highway with one open lane and a highway with four.
Part 6: What This Means, Next to Mojo's Story
Put my last two posts side by side, and a pattern shows up. Mojo tried the loud way build something new, promise everything, end up leaving Python behind. Python tried the quiet way thirty years of patience, one engineer, one careful fix, no new language required.
Maybe that's the real lesson hiding underneath both of these stories: the loudest, boldest fix isn't always the one that survives. Sometimes it's the slow, quiet one that actually holds.
I'm still just weeks into this world of code and languages. But if there's one thing I'm starting to believe, it's this: the biggest problems don't always need a brand-new answer. Sometimes they just need someone patient enough to fix what's already there.
Top comments (0)