Dio9sys

@da_667
RIGHT

like, I'm sorry, but if Palantir is telling me to use something, that is a pretty good sign that I should not be using it. If Palantir made their own Linux distro and it got popular then I would switch to BSD just to keep alex karp from haunting me.

Every time I am shown a new AI thing, even if it makes me go "hm, neat!" I cannot help but weigh that output against the cost, or wonder why an AI needed to do it when existing tools can already do it. Like, do I really need an LLM to tell me how to use the python sockets library, or do they have a readthedocs page that has already existed for years? Does the AI spending $150 to theoretically get a shell on quickJS perform much better than fuzzing? Does that picture of a dog on facebook justify the rise in RAM costs?

It seems like every other thing we are told to have media literacy, and compare the pros and cons and compare against alternatives...unless it's AI, in the which case we need to accept everything people say about it in a vacuum and immediately outsource our thinking to it out of some fear of being "left behind".

Jan 20 at 2:10amWeb
Show threademilianosandriJan 20
@Dio9sys @da_667 Totally agree, lately I'm wondering how much of the repetitive code that AI boosters brags to have automated is actually trivial stuff that slightly smart programmer were already able to automate with much simpler method who doesn't require a pile of energy hungry GPUs. I'm talking about stuff such as: IDE macros, snippet engines, code generators...
If I need to do repetitive stuff such as converting a sample JSON into a class for my language of choice I do it with few clicks, no need to wait a minute for an LLM to do his thing, I take care of repetitive edits with Neovim macros and lately I was surprised by how Symfony CLI was able to initialize a fully functioning authentication system with a single command, no "AI" needed.
Show threadDio9sysJan 20

@emilianosandri

@da_667

It's like we're reinventing boilerplate code from first principles