LLM, like fire is a powerful tool. Some people play with fire and achieved great things, some play with fire and got burned. A number of them achieved great things and got burned. We need to understand that and learn from our mistakes.

The way of thinking it is, telling Claude to tackle the problem 3 times, each time it may or may not use different approach, fix or improve on things it did previously.

This slowness is mostly because of OLTP features

If you want the agent to be more aware of the tool, prompt it to create a skill for whatever bash commands you wish

If you're interested, check EEVblog's video on this

That said, they also use them as slave labors.

Maybe that's what ICE is going to do with the plan to setup large detention centers in the US

Quantization the better approach in most cases, unless you want to for instance create hybrid models ie. distilling from here and there.

It is unreasonable to expect pelicans to ride human bikes, they have different anatomy.

Just like subjects in other fascist regimes, be damned if they do comply, be damned if they don’t.

I see this a lot here

That’s the conventional view. I think there’s another angle: train a local model to act as an information agent. It could “realize” that, yeah, it’s a small model with limited knowledge, but it knows how to fetch the right data. Then you hook it up to a database and let it do the heavy lifting.

That said, open models are not far behind SOTA, less than 9 months gap.

If what you're asking about those models that you can run on retail GPUs, then they're a couple years behind. They're "hobby" grade.

What's likely is that there won't be anything open / significant coming out from them anymore