Even my Linux PC talks to me through the terminal. It is just that it is very specific and technical but at the same time isn’t invasive and hogging my resources.
It is pretty cool that you can run an LLM in your terminal and generate valid tar or ffmpeg commands with natural language.
I practically live in the terminal, I’ve never seen an IDE that compares.
I’ve been working on my own agent harness/cli app for months and being able to run local models on my hardware, running my own software, in my preferred dev environment is great. 100% private, with nearly zero security risk.
I’ve got a framework desktop for this so it’s running 10W to 80W peak during generation, which is about two incandescent lamp sized light bulbs when in use. That’s totally disregarding the model training though.
I’m trying the new olma open models too, so hopefully they get better and I can have better accountability on the training side too.
I’m pretty opposed to the waste that Meta et all are generating for this stuff.
Even my Linux PC talks to me through the terminal. It is just that it is very specific and technical but at the same time isn’t invasive and hogging my resources.
It is pretty cool that you can run an LLM in your terminal and generate valid tar or ffmpeg commands with natural language.
I practically live in the terminal, I’ve never seen an IDE that compares.
I’ve been working on my own agent harness/cli app for months and being able to run local models on my hardware, running my own software, in my preferred dev environment is great. 100% private, with nearly zero security risk.
I’ve got a framework desktop for this so it’s running 10W to 80W peak during generation, which is about two incandescent lamp sized light bulbs when in use. That’s totally disregarding the model training though.
I’m trying the new olma open models too, so hopefully they get better and I can have better accountability on the training side too.
I’m pretty opposed to the waste that Meta et all are generating for this stuff.