I have smart home things but they’re all locally driven by HomeAssistant.

I want the convenience of the automations without ever having to worry that some corporation can or will take the functionality from me.

This is why my personal use of AI has been focused pretty cleanly on “doing what I already do, more thoroughly” - By not turning it into a “ship more code more faster” machine, it’s a “can explore my code and answer questions and help design things more thoroughly” machine.

I tend to go with “AI-augmented” development because I’m shipping the same things I’ve been shipping - Just with a way to quickly brainstorm and compare ideas on something my team members may not have time for. I can propose my ideas and have some LLM tell me what the downsides of my approach would be - or what I should guard against.

It’s crazy to me that folks are treating them like sources of truth when they should just be an untrustworthy second opinion that is faster than you. I think of it as an intern with speed but questionable taste lol.

The sidebar, in its entirety. Snapshotting/undo-that-also-undoes-filesystem-ops. Web interface. Easy model-swapping and provider selection. Workspaces (can be achieved with Claude Code and git worktrees).

There are probably more but I’m on mobile and can’t think of them off hand.

Opencode does way more than Claude Code does

Never had this issue with neovim :3

Arch since 2018, mint for a few years before that, Ubuntu from 2008-2014

The fourth half is caching

In your defense, I’ve thought the same joke every time I’ve seen it lol

Funny to see someone else with an active distaste for his videos. He sets off predatory alarm bells in my head and feels smarmy to me.

I used Mopidy for NCMPCPP for Spotify for a few years, that was fun.

I just run the llama-swap docker container with a config file mounted, set to listen for config changes so I don’t have to restart it to add new models. I don’t have a guide besides the README for llama-swap.

Especially from a 7b model

run a local LLM like Claude!

Look inside

“Run ollama”

Ollama will almost always be slower than running vllm or llama.cpp, nobody should be suggesting it for anything agentic. On most consumer hardware, the availability of llama.cpp’s --cpu-moe flag alone is absurdly good and worth the effort to familiarize yourself with llamacpp instead of ollama.

As someone with dark hair and quite pale skin, I was basically a perfect case for it. At 7 sessions so far, the number of hairs that still grow on my face are countable - 43 of them by my count - I still do have to shave but it takes like 30 seconds because they’re all focused on my upper lip and one particularly stubborn spot on my chin. The important part is that when I shave, I have no beard shadow at all now. I may not 100% pass yet without makeup but it makes makeup way easier to not have to also collor-correct for beard shadow.

Others have answered with more detail but yeah, it also does take testosterone suppression or you’ll have to go back every so often to address any new/rejuvenated hairs. A cis guy friend of mine has to go just about every year.

I’ve been in the industry for 13 years, a technologist using Linux for 19 years - I think I’d count as a greybeard if I hadn’t lasered it off as part of my transition lmao

Yeah I’ve been looking at cachy for a while, the main thing is jumping distro is a pain lol

Oh wow. Maybe it’s time I move to another distro after 8 years

Only if they’re contributing through GitHub and not through local AI coding apps like Opencode or Claude CLI

Someone forgot a debounce

And then made no effort to proofread it