Someone who uses Kubernetes on a daily basis here.
The visualisation thing looks like a nice to have feature, but nothing a K8s administrator can't live without. What's really interesting is the yaml generator tool. This is something that would significantly improve my productivity if only it were a little bit more uses friendly. Building new deployments/jobs/resources in general involves a lot of 'kubectl explain' invocations and documentation searching, not to mention the painful experience of making sure the yaml is properly indented. This tool has a lot of potential when it comes to improving that experience, but it would have to be simpler and faster, e.g. remove the popup windows after clicking "plus" and instead have a panel at the bottom of the screen. Do something to remove the extra clicks when adding nested items, maybe a tree-list structure with quick-add option? I mean, after you get accustomed with all the fields you don't even need the description anymore, you just want to add new attributes and assign values to them.
The way I've saved myself from YAML insanity in k8s is by using kubecfg  and kube.libsonnet . I can now build abstractions that work for my particular deployments and never have to worry about extraneous verbosity in day-to-day operations.
These are very good suggestions, thank you! The editor was the first chunk of Konstellate that we built, and we definitely want to keep adding to it. We have the nested feature on our roadmap, but I hadn't thought about the inline field addition. I will let you know when we have some updates. Thank again for your feedback it's incredibly helpful!
Developer here. I suppose the simple answer is that this was a personal/side project at my job and I picked the language I most enjoy. Additionally, I've been hacking on a couple of libraries for writing web apps for the last few years and wanted the chance to hone them on something non-trivial.
Clojure is very pleasant to work with and is exceptionally useful for rapid development. My colleagues at JUXT certainly would not have been able to build our new bitemporal Datalog database  without it!