Low Code vs. High Code
Palantir's pipeline builder and Microsoft's power automate flow have the following things in common. They both are extremely unwieldy and have a high barrier to entry. This could be solved by a very well versed MCP server but Copilot is miserable and I haven't tried AIP yet.
I think the appeal of a flow modality is simulating a compiled language which is an incredible concept. You still get control flow and switching (the basics of higher level system design). You get syntax errors in the form of flow checkers and semantic ones just go completely under the radar like usual.
The case for low code environments is to make automation more accessible to non programmers while providing safety checks. Power automate for example shows each code block in raw JSON (it's a series of API calls) but will not allow you to write the block in pure JSON. You must write it by dragging and dropping and configuring parameters.
Palantir's system is considerably more complex because you have to establish objects and relationships between them before you access them. It's like configuring every variable and connection before you write a single line of code. A pretty unintuitive way to rapidly prototype. You've also got to create the memory backings for certain objects like linking them to sheets etc. It's like the highest abstraction layer for C++ imaginable that manages to keep a large percentage of its complexity.
Sometimes you have to let people do hard things because it's better for them. The obvious phrase here is to teach a man how to code gets them much more fish than teaching him how to fish.
Post writing notes: Found this interesting bit: https://jameelur.com/blog/overcoming-friction-leads-to-growth