Why aren’t more startups utilizing visual node editors like Blueprint from Unreal Engine or Blender’s node editor?
#VisualNodeEditor #Startups #WorkflowAutomation
Have you heard of Sequence, a startup offering a node editor for personal finances? It’s an intriguing concept that could revolutionize how users interact with complex data.
##Benefits of Node Editors:
– Automation for non-coders
– High customization without code
##Challenges:
– Complexity for onboarding users
What other industries could benefit from node editors? Perhaps stock trading for building trading strategies or legal compliance, with potential integration like Alpaca for simplification.
#IndustryApplications #WorkflowSimplification #NodeEditorUsage
Let’s discuss why visual node editor-based startups aren’t more common and explore potential opportunities for growth in this innovative technology.
This kind of logical/structured thinking is not actually a skill that most people have. The people who can do it want more control than a visual editor gives you and the people who can’t do it aren’t given any new powers by having a visual interface like this.
I also think it’s very rare for visual editing to be the core value prop. I think what you need is a valuable offering where visual editing is a feature, or editing is done in code with a visualization layer that you can use to view the defined behavior. If managing the flow of my money isn’t a problem I have then the node editor doesn’t magically make it a valuable tool for me.
There is buildship.app for backend development. I am not in any way affiliated
Node-and-wire UIs are fine for basic no-code applications but tend to devolve into a jumbled mess of spaghetti when trying to do anything complicated. At that point they need to be rewritten into a traditional programming language, which may cost more than doing it that way in the first place. ComfyUI has seen some success with the paradigm recently, though, for AI image generation workflows. I think it’s best suited for cases like that where a resource is getting transformed/worked on, without a lot of conditionals or loops.
I think I fell for this mistake a few times, and it’s the difference between power users, casual users, and visual power users.
Visual editors are great for empowering casuals to compete with power users, or for power users to clear tedious work, but those visual editors often struggle to keep up with the required customizations.
UE5 / blender etc. are targeting a visual power user so it’s a rare combo that makes visual editing their optimal tool. Not like accountants (excel) or developers (IDEs).
I wanted to build a node editor for my startup where I’m beautiful building branded email sequences for firms to communicate with their clients, I was considering a node editor approach similar to lemlist. But sadly, email flows with known clients are pretty standard, so it’s a bit overkill for: initial email -> reminder emails (frequency can change) -> possible edit/revision/review emails -> completion email.
If you want an example of a great node editor, I think lemlist (a cold email outreach platform) did an amazing job given how hostile cold out reach is and the various tricks they have to use to warm up leads.
I have considered using this type of UI multiple times and generally decided against it. There are a number of reasons why I think it’s generally not that great:
* Lots of thing are more linear than you might expect, making it linear makes it much easier to follow, even an if/else can often be presented more linearly
* Quite finicky generally, not that quick to input things, then messing around with the connections etc.
* Easy to make mistakes: Especially with many nodes it can get quite messy and hard to see what is what, and how things flow, spaghetti basically.
* Steep learning curve: Basic users will not be able to use it, requires effort to learn
* Hard to get right: the basics are fairly simple to make, but to make it easy to use is very difficult
All in all I think it works if the use case is complex, it’s visual programming for highly domain skilled non coders. It might not have the syntax of programming but with enough nodes it has a lot of the complexity. Often it is a code alternative (this applies to blender, unreal, also max for live, davinci resolve)
If you do it, consider locking the nodes (not free drag and drop) and picking one main axis (top to bottom, or left to right). Look at Zapier, Apple Shortcuts, and similar, even filters in excel/numbers/google sheet. It’s still complex, but a lot easier to deal with.
To summarize: Do you have a very complex use case targeting tech savy non coding professionals, then it might work. Otherwise there is probably a better way.
https://camunda.com is pretty popular in enterprise
Node editors are cool, just niche. Adoption’s the real challenge. How’s integration for AI workflow platforms?
Because all the big tech cloud offerings already offer low code / no code UI’s now. Not many people are interested in competing directly against Google, Microsoft, Oracle, etc. with their personal finances on the line.
Nodes are just another interface. A good one at that but it has nothing to do with a particular startup or product. Anyone can implement such an interface and many do. But it is not at all the breakthrough that you might imagine it is.
Rhino/Grasshopper and Labview are probably the two ones that come most to mind for me in the engineering space.
I’ll say this, as someone who worked on quality engineered projects like Aerospace or nuclear we had the HARDEST time qualifying and QC checking visual editors and eventually banned labview at one facility. There often wasn’t a way to print out or archive information for auditing purposes, rollback etc. It was much harder to see changes or detect logic errors compared to raw code. On more complex projects you invariably end up looking at bowls of spaghetti with overlapping traces trying to figure out how it worked.
I think they are great for fixed workscopes and have the advantages you seek but I think a lot of the sentiment is either “learn to code/hire coders” or “this is fine for prototyping/low risk processes, but I can’t build a reliable production system on this”.
Power users love node interfaces. Regular people hate node interfaces.
The average person does not think in flowcharts and that may be very hard to understand for technical people.
I have worked for many years on Apache NiFi, which is a very popular interface like this for live data processing and movement.
Stone of the challenges are handling parallelism and reuse of flows, and the complexity scaling of large implementations. Other comments in this thread hit on the problem that low-code tools are proliferate and FaaS solutions mean larger Frameworks, particularly in the cloud, aren’t as necessary as they used to be