i mean python is 99% backwards compatible so as long as you tell your tooling you’re working with 3.10 it will warn you about using stuff that’s too new. that’s why the shipping version is usually enough. in general it’s not recommended to have multiple versions of python3 installed at the same time, but if you are a habitual venv user it’s usually not a problem. however i have also run into the issue of some versions being “too new” for a project, where the thing just would not work with newer versions.

basically, if your issue is only that you don’t want to “contaminate” an older codebase, that can be solved by configuring your tooling. but if your issue is that the thing just doesn’t work with the “wrong” version, you’re probably best of using a container. a user installation of the version you want will work but having multiple installations is annoying.

use something like uv to install whatever python version you want and set up a venv with it. that way you can have project-specific python versions.

i feel like this should take into account the month when the program was compiled, or at least when its process started.

to feel more like an actual book probably

it’s quick, it’s easy and it’s free

i mean i haven’t signed anything…

we’re in web 3.0 now, apis and data access are a thing of the past. so scraping it is!

yeah this project has been on github for six years and seems to have been closed source before that. it’s a graphical automation tool.

like, everything can be used with ai. github itself has “ai agent” plastered everywhere. it’s just a buzzword. doesn’t mean it’s built specifically for ai.

could be one of those cases where the product predates ai but some c-level asked an engineer “could we use this for ai” and the engineer said “i mean, technically yes” and then marketing changed every single mention of the product

i like how everyone got hooked on the cgnat thing when i gave the actual solution in the main post. but yeah there’s always the option of not doing anything until i see issues.

i’ll worry about the nat traversal when i get my bouncer back up, but it will probably be less full-featured than pangolin. previously i just used a reverse ssh setup but that was a bit too rudimentary.

that’s also a possibility, but i’m going to have to whine to my isp.

as i said i’m getting my bouncer server set back up next year after the datacenter it’s in has finished renovations, so actually getting a public address is not the biggest issue.

i was sort of asking the opposite question to this answer, i think.

my registrar provides ddns, but how does that help with cgnat when thousands of people potentially have the same address?

i’ve set up servers with static ips in datacenter settings before. the way you know you’re online is usually that your cpu activity jumps a few percent from all the incoming ssh traffic from russia and china. i don’t want to risk anything happening to my home server.

this looks like uiua!

last i looked at what talon requires, their unresolved asks spanned 17 years of wayland development.

all of the issues listed are closed so any recent version is fine.

also, you probably don’t need to deploy this unless you have a problem with bots.

pytorch is a unique kind of mess