😷 The JS tumor is out of recession again... 😷

# [ $davids.sh ] · message #287

😷 The JS tumor is out of recession again... 😷

Prediction

• JS will see its biggest growth ever in 2025
• For developers and studios specializing in full-stack apps built with neural networks, JS will become the primary stack
• Entering the frontend market will become harder, and prices for pure frontend devs will start dropping overall

Why?

• Most code in repositories? JS
• Who uses JS the most? Startup founders
• What are the hottest startups right now? AI

As a result, besides the organic growth of JS developers (easiest entry point with high salaries, tons of courses, full-stack opportunities), AI now understands JS better than anything else and is focused on it.

And I’m not talking about "knowing JS as a language"—no, neural networks handle other languages fine. I’m talking about the ecosystem of libraries and SDKs, which AI is much more familiar with in JS because they’re so widely used.

Cursor is evolving at breakneck speed, Replit, Lovable, and other full-stack app builders are gaining traction.

Bottom line: JS isn’t going anywhere for the next (my life expectancy minus current age) years. Good thing AI writes in TypeScript, so it’ll remain the de facto standard too.

#js #ai

• @ [ $davids.sh ] · # 1846

  By "unreadable fronts," I mean those who (1) are prompt engineers even more than front-end developers, and (2) are full-stack developers capable of taking responsibility for both the front-end and everything needed on the back-end for the front-end.

• @ Nikita · # 1847

  The same can be said about Python then

• @ [ $davids.sh ] · # 1848

  Actually, I just thought that the entry barrier and the volume of vacancies for all web positions will change significantly, but this is an idea for a separate post.

• @ Nikita · # 1849

  Waiting for the post

• @ Daniil Tarasenko · # 1850

  Well, if we're talking about Claude, it initially prefers Python for most tasks. If you start talking to it about fullstack, it will probably suggest JS, but if you don't mention the stack at all, it most often suggests Python for the backend.

• @ Nikita · # 1851

  The Gopotа chat is also initially written in Python and executes code in Python as well.

• @ [ $davids.sh ] · # 1854

  Gentlemen, I'm not saying "JS is first in AI" by any means, but it definitely takes a top 3 spot.

• @ Vova hardvair smartvend 🛍️💻 · # 1883

  While things are still somewhat understandable with neural networks on the front end, they write complete nonsense on the hardware. Here's an example of "checking divisibility by 4" in assembly. It's frankly a jumble of letters instead of code (you really only need one line andi r22, 0x03), and it's also incorrect.

  A neural network can be used for TDD, but the convergence of task execution time estimation becomes questionable, and in the end, you need to write many times more tests than code.

• @ Vassiliy ITK Kuzenkov · # 1885

  And what was the initial prompt?

• @ Vova hardvair smartvend 🛍️💻 · # 1886

  Write me a code in avr assembly that check is number (32-bit) divisible by 4

• @ Vassiliy ITK Kuzenkov · # 1888

  You're right, my code is a bit off 🙁

• @ [ $davids.sh ] · # 1889

  (If you're interested, there's a really cool and interesting discussion about AI in the chat comments, so check it out)

  #nativeintegrationofbullshit

• @ Vova hardvair smartvend 🛍️💻 · # 1890

  Also incorrect, but this time it's an optimization issue. Why use tst if andi already sets the flags?

• @ Vassiliy ITK Kuzenkov · # 1891

  Yes, it can be cleaned up, these checks are clearly unnecessary, but there's a comment there, it's easy to figure out what needs to be deleted.

• @ Vova hardvair smartvend 🛍️💻 · # 1892

  That's exactly it, you have to think of it. "June" will nail it.

• @ Vova hardvair smartvend 🛍️💻 · # 1894

  I teach a little, and when I see this kind of crap, it's a code marker from a neural network.

• @ [ $davids.sh ] · # 1895

  In short, yes, we're sliding back into mandatory human validation again.

  And if we compare it to the industrial revolution, we've kind of already gone through it: yes, machines can and will replace line personnel, but the engineers who develop and maintain them are and will be needed.

  So it is with AI – we are these engineers.

  I would hypothesize that it's not "programmers are no longer needed," but rather that there will be a shift in requirements: mid-levels will need the knowledge of current seniors, juniors will need the knowledge of mid-levels, and current juniors will be screwed.

• @ Vassiliy ITK Kuzenkov · # 1896

  I have the same experience. I really ask people to think before making decisions based on neural networks.

  Btw, I'll check this out next time at a junior interview.

• @ Vova hardvair smartvend 🛍️💻 · # 1897

  Middles remain middles. Without real seniors, companies slide first to "startup" status, then "studio," and then they go to the poorhouse.

• @ Ivan ITK 🚫 · # 1898

  My favorite example for all of this is: if a senior developer comes to you who has been writing the same thing for 10 years, AI will replace them first. But if a junior developer comes to you who has tried more different technologies in their year than a typical mid-level developer, AI won't be able to surpass them in the foreseeable future, for sure.