[datsci_est_2015]

The better your code is architected, the less powerful model you’ll need for it to make sense of it.

E.g. a well-designed deployment (infrastructure-as-code) repository doesn’t need a frontier model to be understood well-enough to create a new job / service using sibling jobs / services as templates.

And this already saves me dozens of minutes per week, although it’s not a 2x multiplier in my efficiency.

[varispeed]

The issue is that local models are dumb and tend to make mistakes than look good at a first glance. So any "saving" is quickly ruined by having to do an extensive review. You might as well just write things yourself.

[datsci_est_2015]

I use it as code scaffolding, which means in a way I’m often rewriting it. For me, writing from scratch isn’t the same amount of effort as using a code scaffolding tool.

[lukaslalinsky]

I disagree, even though I'd love for it to be different. With models like Opus, I can give it a good architecture and expect good results. For many of the less expensive models, that is not the case, they make mistakes, you need to over specify, they get stuck in a loop, etc. As you get to the models you can realistically run locally, it gets so frustrating I'd rather be writing the code myself.

[datsci_est_2015]

At what point will local inference catch up to today’s cloud inference? Will it ever? If it doesn’t, does that imply a certain dead-end for the LLM inference industry?

[lukaslalinsky]

I don't think at any point in foreseeable future we will have terabytes of RAM for dedicated LLM chips at home.