It is pretty clear that today’s AI, needs a visual programming language.

The AI needs to live in the nodes, taking programs to a while new level.

It is also rather clear that user interfaces are trouble, nobody uses X/Y coordinates anymore.

The layout engines are responsive now, and the only one I could find, is cross-compiled for JavaScript.

So things are pretty silly, responsive user interfaces will need to go as well.

Frankly, I hate the idea of a different UI for every website, that is not how computers are meant to work.

Crazy designs, force your users, to feel stupid, the moment they sit down to your website.

Feeling stupid, what I experienced, on important government websites.

I dare not to learn how to shop in such a way, that the store delivers stuff to my trunk.

I hate it, I hate the idea of needing to learn how to do that.

Of course, AI, is just magically smart enough, to be able to do the shopping for you, based on a chat.

But that feels flimsy and foamy, and I would worry about getting strange items the AI hallucinated.

I don’t know what the UI should be, though I suspect it is generated based on rows in a database.

So it is very much like Access or FileMaker, because it works, it is rational and efficent.

You design a database structure, and get the UI for free.

If it is a relational structure, then you get multiple auto-generated paned in your database.

Tightly hugging those rows and columns, creates a UI for free, and that is better than manual work.

The layout collapses for mobile, but the best way to “see it” is to think about the Metro squares.

Automatic layout, so simple, it hurts.

It is better to have the user think for a second, then to have them feel stupid an lost.

Here is the big idea, a man in the middle HTTPS proxy, that forges all the certificates and locally caches all the data.

Then the data is served from cache, under a uniform user interface.

There is an AI program that will help with odd proxying tasks, won’t fix government websites, nothing can fix those.

But it will fix manifest 3, and 4, 5, and more and more restrictive web.

The user interface for that, has to be as adaptive, as quickly everything will change.

And that means a visual programming language, Blender geometry nodes is the baseline and the workflow.

It is not as hard as it looks, I programmed Signalcraft, a reference implementation in about a month.

It is just SVG and Reactive Object and Arrays, under a simple OOP Application Architecture.

I made my holiday into a Hackathon, and created a Nostromo theme inspired by old Retro-futuristic graphics.

I dare make atoner bold statement in closing, there is nothing as fun and rewarding as programming a Visual Programming Language.

But I must admit, this being a hackathon with an 11PM deadline, it got me a few times, as I had to hurry.

I finally finished a few minutes late adding project templates, and somewhat fully functional loading and saving of files.

But there was no time to add pretty error messages, even though I have toasts, I had to use the old alert box to remind the users to select a node to execute.