Hacker News new | ask | show | jobs
by _pdp_ 409 days ago
I just want to make an observation based from previous experiences.

These styles of programming environments are not new. We have seen many such tools in the past. I think they are useful (to an extent) when dealing with high-level compositions such as orchestrating a video pipeline, shaders, simple logic.

They start to fail very quick when you need to implement more advanced algorithms and processes that are not yet abstracted in reusable blocks. Needless to say, it is easier to write code at this point. Just less cognitive load and frankly the visual model does not help when you have a solid mental model - you can keep more information in your head in abstract forms than the eyes can scan and process.

I am lurking in the n8n community and I have noticed similar patterns. The more elaborate workflows are just not as robust as they could be if you write code - and they happen to be very difficult to maintain too and full of bugs.

Other than that... these types of framework have their place and as I said, overall I think they are useful as long as they are used in the right way.

2 comments

It’s all about using the right tool for the job. In control systems I have seen nightmares of for loops and if then conditions replaced by a skilled practitioner to a beautiful block diagram that solved the problem and illustrated to the reader the form of the solution.
"The past is never dead. It's not even past" - William Faulkner

Visual code is all over the place if you look; Any time something advertises itself as "low code" or "no code" it's a drag-and-drop flow based UI like this one. Most (if not all) ETL systems work this way, and all kinds of low-code AI Agent frameworks are cropping up.

I do agree that there very high prices to be paid for this stuff, and worst of all most of those costs are hard to see, and only crop up late in the project, when, as you point out, the more complex algorithms and processes crop up.

https://en.wikipedia.org/wiki/Extract,_transform,_load https://nifi.apache.org/ https://hop.apache.org/