Hacker News Reader

Comment by conartist6 | Hacker News Reader

Feels like a thick layer of AI-generated glaze on what's basically just another no-code platform... except this one with the innovation that it's actually a book that you buy?

While the copy sounds like complete fluff-n-puff, I honestly believe almost all of it is right, both about what's coming for the future of coding and the role that visual "lego" programming will play in it.

There's just one really major thing the author got wrong: the reason visual programming hasn't (yet) seen broader adoption. The reason is that each visual programming tool consists of one language married to one editor. Who wants to re-learn a whole new editor each time they need to work in a different language?

View on HN

Coupled languages and programming tools had OK adoption back in the day: Turbo Pascal, Visual Basic, HyperCard. But times have changed, perhaps in no small part thanks to the open source movement.

I used LabVIEW heavily for a few years in the mid 90s. What killed it for me was the sheer physical labor of programming in a graphical environment. All of that detailed mouse work gave me severe eyestrain headaches and wrist fatigue. When I'm typing text, I'm barely looking at the screen. Granted, this may be a personal physical limitation, so I can't offer it as a general argument against graphical programming.

Another issue is that decoupling the tools and the language lets you experiment with new languages and features with a bare minimum of effort. I've even tried out my own toy languages. But building a graphical editor, even a general purpose one that could serve multiple graphical languages, seems like a boatload of work.

The core principle of Pipe is that it is not trying to replace text-based languages. There will always be areas where text is better. What Pipe is doing is visualizing structures on top of textual languages. It means that mouse manipulations in Pipe are going to be significantly less intensive because Pipe does not try to implement x=x+1 visually (I think it is the key mistake in majority of existing VPLs including LabVIEW and hence physical fatigue). What Pipe does is encapsulating non-visual code that does x=x+1 into a reusable visual element.

Also, AI can generate textual code calculating x=x+1. It means we mostly do not even need to touch textual code - AI will do it for us perfectly. All we need to do is converting AI-generated code into visual components for composing visual workflows.

To sum up, AI and visual programming create a perfect synergy: AI generates code and visual programming composes generated fragments into visual components. This should minimize physical efforts needed to manipulate visual workflows as devs are going to work only on high (visual) level of abstractions while AI will be providing low (text-based) abstractions.

Reply on news.ycombinator.com