The Turing Award lecture usually sucks. Formal, banal, safe. Except when the giants of CS decide to break the script. John Backus used his to launch functional programming. Ken Thompson warned us that our compilers might be lying to us. Edsger Dijkstra told us to be humble before our own brains.
Kenneth Iverson did something different.
In 1979, he gave us “Notation as a Tool Thought.” He argued that how we write code changes how we think. A good notation frees the brain from unnecessary labor. Iverson created APL for this. APL looks spooky, dense with symbols that look like Greek letters on acid. But it allowed mathematicians to think in equations, not loops. No more translating mental math into clumsy Fortran. It worked. It wasn’t adopted everywhere. It didn’t matter. It proved the point: two languages can become one.
Now, sixty years later, we have the Two-Language Problem.
Python rules. Everyone loves it. It’s the friend at the party who tells you everything but drives terribly. Python is slow. Brutally so. Defenders wave their hands. It doesn’t help.
Researchers prototype in Python because it’s friendly. Then, when performance matters, they rewrite critical sections in C++ or Rust. They do this twice. AI coding agents can’t fix it. Optimization has a floor. If the floor is slow, the building stands slow.
Think about construction. Wood is easy to cut. You can nail together a shed on a Saturday. Steel is hard to work. But you don’t build skyscrapers with a saw. What if you had wood that was strong as steel? Or a language ergonomic as Python but fast as C?
We are greedy.
Four computer scientists wrote those words in 2012. They were Matlab users. Lisp hackers. Pythonistas. Rubyists. Perl junkies. They hated that every tool they loved was perfect for something and terrible for something else. They wanted it all. Open source. Simple for beginners. Powerful enough for the hardest hackers. They called it Julia.
I met Julia in 2017. It was a fluke. I was listening to Sebastian Seung, a neurosciententist mapping brain connectomes. The name itself felt like an apology for the rest of the field. Forget C++ or MUMPS. Or even Haskell. Julia was winsome. Simple.
It was also carefully engineered.
The creators had watched other languages fail. They took the good ideas, left the baggage. By 2026, the Julia community is weirdly mature. No drama. No flame wars over syntax. It attracts scientists, not just coders. They aren’t interested in intellectual gamesmanship. They want results. At JuliaCon, people brag about rewriting MATLAB code and seeing 60X speedups benchmarks. Ten thousand times faster than Python, some say.
So why isn’t it everywhere?
Why doesn’t Stack Overflow love it? Why did Python win?
Ecosystem. Python has libraries for everything. You need to do something obscure? There is a Python package for it. Julia doesn’t have that weight.
Corporate patronage. Objective-C rode Apple’s coat. Kotlin rode Google’s. Julia got nothing from the Big Tech giants. It grew on its own.
Or maybe nothing went wrong at all.
Julia is niche. Small. Beloved. It runs the big machines at CERN and NASA. It helps design drugs. It’s doing the work.
Do you think any language solves the Two-Language Problem eventually?
I doubt it. The split is inherent to software. Games use C++ engines and Lua scripts. Server backends mix Python logic with Rust performance. Frontend development fails every time it tries to adopt Rust or Go.
We keep wanting the magic bullet. Wood that doesn’t rot. Python that doesn’t stall.
We build with what works. We switch tools when it doesn’t. That’s not a bug.
It’s the job.
