Python Is Very Slow. Can Julia Solve the Bilingual Problem?

As a nation, The “reward issue” is formal and restrictive. But there is one very interesting thing about this law—the speeches given by top computer scientists at their event. Turing Awards.

Some read like manifestos: John Backus’ “Can Software Be Released From Von Neumann Style?” (1977) inspired a new theory that led to functional languages ​​like Haskell. Some caveats:Thoughts on the Trust Trust” (1984), Ken Thompson pointed out the dangers of closed-back compilers, possibly preventing security problems. Edsger Dijkstra, inThe Humble Programmer” (1972), exhorted his followers to be alert and tactful and to accept “the inner weaknesses of the human mind.”

For our purposes, consider Kenneth Iverson’s 1979 essay, “Cognition as a Tool of Thought.” In it, he showed that mathematical notation is not only useful – CO₂ of carbon dioxide, 3,888 of MMMDCCCLXXXVIII—makes new information readily available. As the mathematician Alfred North Whitehead once said: “By ridding the brain of all unnecessary work, a good word frees it to concentrate on higher problems.”

Iverson won his Turing Award for APL, a visual language that began life as a way to communicate between languages. In the early days of computer science, programmers had to think in one language (mathematical notation) but then another program (for example, Fortran). APL was designed so that intuitive functions can be written in parallel as equations – lines of code collapsed into multiple symbols such as + or x. APL turned out to be more powerful than imitation, but it didn’t matter: It showed that two languages ​​could be merged into one.

The year is 2026 60 years since the beginning of APL, and a new type of bilingual problem bedevils the field of computer science. The main programming language is Pythonbut you rule not as a mighty conqueror but as a stealthy king. Python, in other words, is very slow – a fault that even its staunchest defenders cannot deny.

Hence the problem of two languages: Researchers show slow, friendly Python but, for the most important parts, rewrite fast, unfriendly languages ​​like C++ or Rust. This cannot be solved by running around a bunch of AI coding agents, because no matter how slow you make the language, the fastest will outrun it.

These commercial products are available in other regions. You could say that construction, for example, has a two-pronged problem. Wood is a brilliant material for painting, even an uneducated person can see and nail a functional building. But it is not good to build a skyscraper. This raises the obvious question: What if there was something as flexible as wood but as strong as steel? What if there was a language as ergonomic as Python but as fast as C?

In 2012, four computer scientists with strong mathematical skills came together to solve the modern bilingual problem. Briefly called “Why We Created Julia,” they said they started the project “because we are greedy.” Their words start as valentines to programming languages:

We are the power users of Matlab. Some of us are Lisp hackers. Some are Pythonistas, some are Rubyists, some are Perl hackers… C is our desert programming language.

But these languages, they wrote, “are very good for some parts of the work and bad for others.” Greedy as they were, they wanted “a language that is open source, with a free license … Something dirt easy to learn, but keeps hackers very happy.” Julia would be the one language to unite them all.