The last couple of years have seen more changes in programming than possibly the previous 25 years. AI is the reason, and it has turned programming from a skill that took years to master to one that pretty much anybody can do just by asking AI to write it.
Despite that, programming languages continue to be developed and they’re covered by several regular software developer surveys (RedMonk, GitHub Octoverse, and Stack Overflow’s annual Developer survey). GitHub Octoverse is based on the languages used in GitHub projects, which probably biases it somewhat, but not as much as the monthly Tiobe Index which is based on web views. This measures what people write about rather than use, as demonstrated by it including Visual Basic in its top 10 whereas on RedMonk’s chart it ranks at No. 54.
All three (RedMonk, Stack Overflow and GitHub) agree that JavaScript is top, with Python second and Java, PHP and C# also ranked amongst the top ten positions varying by survey. Rust, despite only making it to 12th in Stack Overflow and 20th in RedMonk, is one of three languages on GitHub that demonstrate considerable growth this year. Previous Stack Overflow surveys once had a most-wanted technology category, and Rust had topped that for the previous seven years.
I think it’s significant that on Reddit.com, the Rust subreddit now has the third-largest number of members after JavaScript and Python’s subreddits. You can see the list of programming subreddits here.
Splitting by genre
Based on the Stack Overflow survey, web and full-stack/backend development make up 40-45%. Desktop development is less than 5% of all developer roles, and mobile development is around 3%. The rest, which includes just under 50%, includes embedded systems, game development, data engineering, DevOps, AI/ML and system architecture.
Interestingly, the majority of AI development is no longer done solely by dedicated AI/ML engineers. General full-stack, web and mobile engineers are heavily involved in "Applied AI"—integrating existing large language models (LLMs) via APIs into traditional applications.
Rust is the new common language (aka lingua franca)
It used to be C, but if you’re going to have AI generate source code, picking a programming language that is memory safe and executes as fast as C/C++ then Rust is a bit of a no-brainer. You could have AI generate JavaScript (or TypeScript) but generating Rust includes the ability to generate WebAssembly for free, and WebAssembly outperforms JavaScript when CPU speed is the issue.
Using AI to convert applications written in one language into Rust is not just a possibility; it’s already happened. A recent conversion was the Adobe Creative Suite (16 modules) onto the open-source ArtCraft. This was done by the AI looking at user interface screenshots, reading public documentation and observing feature inputs/outputs. It then generated an entirely new codebase from scratch.
There have been alternatives to Photoshop and other Adobe apps before, but this is a feature-by-feature copy, and unlike the original it runs on Linux as well as Mac OS and Windows, as well as in the browser. Whether it is entirely legal is another issue! If you look closely at the 16 ArtCraft apps, you’ll spot Wordcraft, which is a MS Word clone (though not fully tested) and no prizes for guessing what Gridcraft is a clone of!
Likewise, Postgres has been rewritten in Rust (pgrust) and passes 100% of the Postgres regression tests, although it’s only at version 0.4. It claims to be 50% faster than Postgres on transaction workloads and 300x faster than Postgres on analytical workloads.
Why learn Rust if AI can do it all?
Yes, you can have AI create applications- but if you ever need to modify them or fix any syntax bugs, then not knowing the language could be a liability. A couple of years ago I learned Rust by creating a program to do a task and then have AI generate the code and explain it to me. It mostly worked, but I found that it generated deprecated code which the compiler did not like. I expect that to go away eventually, but just be aware of it.
Also, some of the Rust code it generates might be less optimized for Rust because it falls back on patterns it learned from C or C++ instead of writing idiomatic, safe Rust that makes use of ownership and the borrow checker. To recognize when that occurs, you need to know Rust.
Conclusion
I suspect the distinctions between programming languages are becoming less important to the survey creators because of AI. And that makes me wonder if anyone is developing a programming language that is more compatible with AIs; one that is cheaper to use, simpler to create, run and debug. It’s just a theory though!
AI also allows conversion of existing software written in older, less safe languages, particularly C/C++, to Rust for the benefits of additional speed and safety. This was not viable before AI due to the time and cost of employing human developers. Perhaps it’s now time for anyone with older software to think about doing this before someone else does it!