Setting up a python environment for ML work (pytorch + Nvidia) is simpler with Anaconda, it's a pure dependency nightmare doing it with something else.
Counterpoint: conda is a constant, neverending source of pure dependency nightmares here, that consistently costs us a two-digit percentage of velocity. I'm glad it's working for you, but I really wouldn't generalize. Or recommend it to anyone getting into Python. Use uv, new people, for real.
To be fair, both you and GP are correct. Conda is really really painful to maintain, and it's also much much better for python stuff that relies on native libraries (i.e. C/Fortran/Cuda).
Like, I love uv but GDAL (to use a random recent example) is much easier to install and maintain with conda.
Yeah, I remember finding that the case for a while. I can't remember when things clicked and it was fine without. I moved to arch a while back, perhaps then.