I don't understand the point they are trying to make.

It's very often (always?) the case that something general also solves particular problems.

    A sorting algorithm is an implementation of min()

    A parser also is a syntax checker.

    A route planner is a reachability checker. 

    A computer algebra system is a basic arithmetic calculator.

    A general constraint solver is a Soduku hint maker.

Fundamentally I read this article as: We want to build a simpler, dumbed down clone of Mathematica, so we cobbled together the following pieces... We also needed a way to do arithmetic, so we also include a copy of Mathematica to do basic arithmetic.

I took the point as: don't make the LLM the classifier. Use it to turn messy input into useful features, then let a normal model make the actual decision. That gives you thresholds/calibration you can inspect.

What I'm not sure about is how stable those features are when you switch the underlying LLM or model version.