Paradigm change: computing is dead

This is a fantastic story of a machine that was born once with your dad, then again with you, then again with your child: in 1946 ENIAC came to humanity and was able to automate computing for the organizations, then again in 1974 Personal Computers came on the table and allows workers to automatize part of their jobs. They connected together in 2993 with the Web and in 2007 it came to your palm and pocket as a smartphone, and cut the cord with 3G.

It evolved but had the same DNA: CPU, RAM, Disk, code, and files.

Now it's changing. Really changing. It doesn't evolve, it mutates: it's becoming a really different beast, in ANY dimension: it's a paradigm change.

Who (you?) realize that every single component of the computing ecosystem is changing drastically? Think about that :

1) Train on a #$!% language or...?

LLM are trained on available code bases (open source) so most being React and JS or Python (PHP somewhere? Java?) they become good at... managing JS which is actually horrible for LLM because it DOESN'T FIT their logic.

But they can handle/understand much more pure object OOP, like Python, even better Ruby and Rails, which is "opinionated" so less "configuration" to choose and fail at.

So is it better for a LLM to be "good" by

a) a loooOOOoot of training in a BAD language/tech such as JS?

or

b) a few training but in an easy language/stack such as RoR?

JS is dead.

(oh, btw, not the first time, but a good one: https://plainenglish.io/blog/react-and-next-js-is-dead-something-new-is-finally-replacing-it-for-good-c792c48806f6 )

2) How about a self-training?

Let the LLM learn the basic of the stack, try it, code, "see" and test, try again, etc. instead of trained "by others" code bases? Self experience is the best, no? It hard code your brain much better than a million books.

So... what's the easiest stack (RoR?) even with a small code base? (and good books) to learn and train , then... self experience at... SUPER SPEED?

Tadaaaam....!

Introducing Pharo, the modern (french) Smalltalk! It's pure OOP, super clear, the best simplest OOP langage, pure logic. Super easy for a LLM toreason about, no tricks and traps like JS. AND because of its "live coding" and "all in one IDE/language/runtime", there is NO REPL, no code-run-test-debug-rerun etc. long loop, you TOUCH the running code, you fix (and enrich) it ALIVE! That makes it super fast to self-train.

Training is dead.

3) Languages and frameworks are for humans?

How about new ones made BY and FOR AIs? Ask LLMs to create their own stack they handle much more easily from human discussions to CPU binary. We humans went from assembly to low level languages to high level languages to frameworks (to CMS to ERPs to low code to nocode etc.). That's stupid to ask to a higher level (LLMs) to handle low level language (JS...). Either it translate (transpile, compile, call it like you want) directly from human (english chat) to CPU binary, either it? should handles a "framework for LLMs", much easier for them.

Languages are dead.

4) Let's go further in the (very close) future: code or...?

Is it smart or stupid to have a humain brain (smart), helped by an AI (half smart) to code a... code (NOT smart)? A dumb code? Wouldn't be smarter to ask the AI to DRAW the app on the screen, catch the clicks and keyboard entries, change the screen by redrawing a few pixels (or blocks, or HTML, whatever front and back framework he handles directly, and even better, a framework he invents himself, making it easier for LLMs, not for human)? From CPU and RAM, to GPU, to Tensors TPU to Groq LPU, logic and data were separated, not fuse in "all-in-one" chips. So be the "code", be "smartapps", no code anymore.

Code is dead. (CPU too)

5) Okay, so an AI is trained on everything?

It" knows it all"... and can speak about everything, and show, draw everything. Any app? One to rule'em all?

Apps are dead.

6) from CPU to apps, they're all dead?

A new "system" arrives. It's not about "information technologies" but "thinking technologies". Not data centric but logic centric. Computing gave "computers". Now is the age of Thinkputers. (oh, hello IBM, you sold "Thinkpad", sh...! ;-D)

Computer is dead

7) What screens? Keyboards?

We had opera, theater, cinema, then television, desktop, laptop, smartphone... the screen is shrinking again and again... smart glasses are coming but, that's new, though the device is smaller, the field of view is much bigger: you "see" a 4x6 meters screen, and not one but two, and in 3D! Smaller is wider!

Xreal, Rokid, Viture, TCL Nextgear, and...Meta, Samsung+Google are coming.

The laptop is dead. So is the smartphone. Heads up and hands free at last, look ahead!

And after tomorrow... another beast is already coming:

"Computing" is not dead, but now is the age of "reasoning" with a totally different kind of machines.

And I'm not sure if CS engineers, who have spent nights and years studying, hacking, crafting codes by the lines, are the best to manage those new beasts, think this new era, "train" and not "code" every single details of the machine behavior.

Maybe "professors" fit the need better, with educational sciences, pedagogy?