Alternative views for Squiggle code

[Hazelfire]

I've had some people interested in alternative views of Squiggle models. Particularly:

• Viewing Squiggle code as Collapsable LaTeX formulas
• Viewing Squiggle dependencies as a tree / DAG

Often the idea is that "I would like the entire model on one screen so that I can see all the dependencies between them and how it works".

I personally am not a particularly large fan of these types of things, because:

1. Models are inherently complicated, and viewing models as "One big thing" often highlights that. Often it's easier and quicker to understand the decomposed model vs the total model
2. Code is a good abstraction level IMO, especially with functions + dictionaries. It does a really good job of displaying the semantics of the model.

I nevertheless include this as a discussion to represent other people's viewpoints.

[OAGr]

Yea, I probably share similar intuitions to you.

I think it would be pretty gnarly to convert a Squiggle codebase to a DAG anytime soon. Maybe with LLMs we could do something, but it would be a while.

LateX formulas could be interesting, but I don't see how it could work too well. Happy to get proposals here.

We are thinking of having Guesstimate use Squiggle, that could be one alternative input format.