It was fun writing this article, not only because it reminded me of problem sets from college, but also to see the potential of no-code tools to be just as powerful as a conventional programming language. Proving another tool, like Zapier for example, can simulate a finite automata, can be left an exercise for the reader.

I wonder how powerful DAG based tools like airflow are, no loops possibly make them less powerful than finite state machines

For a summary: "Visual editors are easy to use and popular, but often lack clear engineering principles in their design. There are a few options to choose from when modeling these editors, including Directed Acyclic Graphs and Finite State Machines, depending on the user requirements and how the entities you transact on are defined. In the case of Laudspeaker, the customer journey was modeled as a finite state machine, with customer actions/attributes as the base set of transitions and messaging steps as the various states."