I'm not associated with Unison Computing and I haven't yet had a chance to use it for a professional project but learning the language and exploring their approach to tooling has been an absolute thrill.
I actually found out about Unison because from my own side projects I came to the conclusion that strongly typed, hash-addressed functions were a super compelling approach to highly modular and maintainable programming - especially in the context of LLM-generated code because refactorings and new function generation require very limited context - something desirable for humans but especially for LLMs with limited context. After digging around for something that did this I found Unison and have now mostly abandoned my own tooling because Unison is so much more mature and has such competent people behind the wheel.
There is a learning curve for sure, not just with the tooling but also the language. It's a challenging language steeped in advanced software engineering principles, but I would 100% rather spend my time honing my fundamental understanding of my craft rather than learning another 20 AWS tools which are going to go out of style in 12 months. After becoming mildly proficient in Unison I feel like I have such a broader understanding of programming in general even though I've been a full time backend coder for 15+ years.
As for the tooling, it does what it needs to and does it well with very competent folks discussing and debating the minutia daily. It's a small team and that keeps them nimble with major improvements taking place each month.
Today I'd say that it excels at microservices, things you might otherwise consider a traditional serverless function for, but gives you way more agility and brevity to tweak the application in a surgical and controlled way which is more aligned with the behavior rather than text files. Something just feels very right about storing the AST as-is and manipulating it more directly.
Tomorrow, as more supporting libraries get built and interfaces to outside of Unison get developed, anything's possible really - I'm personally certain that we'll see some amount of continued shift towards making ASTs the source of truth so I see learning about it and following the software as an investment in myself and my future capabilities regardless of whether the future software ends up being Unison or something like it. Unison is going out of their way to do all the right things even if it's not always the most practical thing given the current corroded state of web engineering in general, so I'm eager to get in on that as much as possible.