Learn Programming With Artificial Intelligence As Soon As Possible

Learn Programming With Artificial Intelligence As Soon As Possible

I continue searching for a way to make AI better at programming, and I am now certain that a better AI is not the way.

I just asked my AI to over every pane of my application, and ask what is missing, what can be added, and if the pane is rational.

Think of a pane as a tab in a complex application, it is a series of tightly nit functions.

In a file pane for example, this should make the AI evaluate the toolbar and buttons.

And check that files and directories can be added, renamed copies pasted cut.

AI being aware of the application it is working on, should also evaluate needs in context of the program.


AIs are held back, this prevents them from breaking stuff, hallucinating things, and consuming everything.

But asking AI to be rational about features, is a sign of times.

There are programmers out there wo will say that AI is bad, that it is a glorified auto-complete, and that it can’t create original things.

But the truth is, programming just became possible this month, up until now you needed an army of contributors.

Many of those contributors turn out to be cranky, pushing a feature too hard will have them split your development team.

And pushing hard is where innovation lives, it is where we create the programs that have not yet been possible.


One of my computer programs is a visual programming language, and to make it work it can’t have loops and if-thens.

This would upset very many people, splitting a workflow, or a dataflow is a solved problem.

But in my opinion, it made visual programming childish.

Splitting a flow, and making a human lose track again, makes a visual programming language into a mess of cable and thus a toy.


Pushing this idea would make my contributors fork the program into two, and everyone would slowly lose interest in both forks.


In the age of AI, we just get on with the show, I ordered my AI to add a custom state of the art code editor and file tabs.

Now, my application does not just open beautiful workflow maps, but also supports programming.

As a result I can give my users program templates, and show them how to sink complexity, loops, and if-elses into actions.

Often it is only 15 lines of code, sometimes 50, but keeping complexity inside actions.

And showing simple flows of actions, saved the program from becoming a toy.

I named this Implicit Loops and Logic, and realized that is what the command line was supposed to be about.

The action or high-level-command is supposed to be a complexity sink, that outputs files, addresses and fifos to make a thing work.

But people placed loops and ifs on top of high level commands, and made it into an unreadable mess that hides bugs for years.


I am allowed to talk to my AI this way, because I divided my system into, custom programming language, custom user interface components.

And only then applications, development of which, enhances the user interface components that benefits all my programs.

Adding a code editor to my visual shell scripting, gave all my other applications, a code editor, if they need it.


Being able to make such decisions means only one thing, real programming was not possible until AI started coding.

Real programming is not about compromise, it is about enhancing our abilities.