I use Pi and Qwen 3.6 27b locally on a 4090 for all my personal projects. I still use Claude for day job work since they pay for it, and my employer expects me to use it. I rarely touch it otherwise.
It's a bit of a contradiction. We understand that AI can be used usefully, and to great effect. But if someone else uses it, it's a potential liability.
I think the issue is, we understand our own usage of it, and respect the boundaries of what's possible and what needs to be done to use these tools properly.
But we don't know how the other guy is using it.
We don't know if they're being responsible, and using it in a safe manner.
If they are: great. But if they aren't, we're opening ourselves up to all kinds of security shenanigans.
It's one of those things where we're only going to be okay with it, if we're the ones using it. But that also means other people will be suspect of our code.
It's really a no win scenario, except for inside each of own little bubbles.
That's how I'm finding I feel, too. It's not that I don't use LLMs myself, but I trust myself enough to know how to filter that data that comes back from them and compare it to actual facts. I don't necessarily know the processing of this information from the other person, to trust that they vet the information properly. So instead of it being helpful to me, it becomes incredibly irritating. Especially because I just know they're going to expect me to take that data at face value quite often. And then it's going to be on me to vet the data. How about you just let me get the data myself and cut out the middleman?
I wonder what's going wrong there? Personally I found compatibility and performance on Linux to be extremely good. And just keeps getting better. And that's not even just me, that's all kinds of benchmarks out there. Sorry to hear that. : ' (
I think once the hype dies down, and things start to normalize, the hate will fade into the background. Instead of "AI", we'll just get "apps with features" again.
Funny, I know quite a few extremely talented programmers who cautiously approached the topic, and found that, with proper use, they've found LLMs to be extremely useful. Just a matter of understanding where the boundaries are, and using them responsibly. It's not a magic genie, it augments their existing skill.
This is why, despite enjoying all of this, I really want to focus on locally hosted models. If we don't host the technology ourselves, we're setting ourselves up for a hard fall down the line.
Until very recently, local models been little more than brittle toys in my experience, if you're trying to use them for coding.
But lately I've been running Pi (minimal coding agent harness) with Gemma4 and Qwen3.6 and I've been blown away by how capable and fast they are compared to other models of their size. (I'm using the biggest that can fit into 24gb, not the smaller ones.) In fact, I don't really need to reach for Claude and friends much of the time (for my use cases at least).