> I also find it takes a lot of mentally energy to get my head around them and makes readability harder.

Why so? They take what was implicit and make it explicit, so I would expect this notation to be far easier to understand than transposes and matmuls. You also get better variable names for axes, etc.

My complaint with Einsum variants in Python is that they are not generic enough -- they are specialised to multiplication & addition whereas generalizing slightly to allow other operators makes the code very expressive and elegant, while allowing you to still compile down to efficient code. For example, the Tullio library in Julia lang is absolutely a breath of fresh air.

For something in the same ballpark that I've found to be pretty ergonomic, there is Xarray: https://docs.xarray.dev/en/stable/index.html

What you do is name all the axes of an array, and then for all array computations, you address the axes by name. The great thing is that for the correctness of any of these operations, you never need to remember what numerical order the axes are in, since you just use the names. (For _performance_, the order can be important, in which case you can easily permute the axes to any order you want and then access the underlying NumPy/CuPy/whatever arrays if you need to.)

Being a non-native and a person who frankly doesn't need this, just passing strings that are to be evaluated seems like the worst case scenario for readability and maintainability.

It keeps the relevant information local to the site of execution and fairly explicit. These operations can have quite a few inputs where you need to keep track of, e.g. the size of every axis of every input, that you would otherwise have to keep track of in your head by knowing what happened during the declaration, or using a lot of matmul/vector operations to do the same thing as one einsum. Couple that with the fact that lots of algorithms are making all kinds of intermediate computation outputs with all kinds of possibly varying shapes, einsum/einops helps a lot with verifying assumptions and letting you annotate what each axis means all throughout.