eris VS xla

What really sucks IMO is that there's no compiler support for exhaustive error checks, the amount of boilerplate for handling errors, no sum types like Rust, horrible stacktraces unless you use something like eris, the ease with which you can ignore errors, all make for really poor error handling as it remains. It is now the #1 challenge that devs report for Go as per their own survey. We might see some improvements on this front at some point and seeing the new errors.Join stuff is giving me hope

Attaching stacktraces to errors is one way to make errors more readable. Go has notoriously unreadable errors and when you log the error, the logging function is now at the top of the stack rather than where the error actually originated. There needs to be a simple function that adds some default wrapping when you return an error because as it is, you need packages like Eris to make the errors even halfway readable

Actually just stumbled over eris which seems decent.

Would it make sense to have a backend support for OpenXLA, Apache TVM, Jittor or other similar to get free GPU, TPU and other accelerators for free ?

GoMLX (github.com/gomlx/gomlx) is a fast and (relatively) easy-to-use set of ML libraries built on top of OpenXLA, a just-in-time compiler of numeric computations for CPU, TPU and GPUs.

I'm doing something that uses XLA to quickly execute computation graphs (in CPU, GPU or TPUs). One builds the graph, and later executes it (quickly) on actual values (sometimes large tensors).