Honestly, I think we could’ve pulled off a lot earlier if GPU development had invested in GPGPU earlier.
I can see it now… the a national lab can run ImageNet, but it takes so many nodes with unobtanium 3dfx stuff that you have to wait 24 hours for a run to be scheduled and completed.
All modern AI tools tend to put `from __future__ import annotations` line into Python code which isn't required by recent Python versions.
I guess this is because they all are trained on an intermediate code bases when this was required some times (even at that time the real usage was quite rare).
And now this line is parroting everywhere like a junk DNA :)
I am not an AI and I missed the memo on this. I put that line in whenever I need to use a forward-declared type annotation. Last I recall reading about it, there were some deep issues that meant it had to stay in __future__ indefinitely. Is there a PEP / release note about it?
I was sure this one was for Python from 3.7 up to 3.10. But it is still here..
But as for 2026 it will not stay forever as it was supposed to - there are 2 PEPs which are suppose to fix it in Python 3.14 - https://peps.python.org/pep-0649/ and https://peps.python.org/pep-0749/
Maybe they could, as ASICs in some laboratories :)