I think going directly local is great, I absolutely love it.
1/ Its just that for me, making the first step and getting started seemed easier when just getting an api key on payperq for 2000sats and play around a little, than to set up a local LLM.
2/ for developing more complex logic that goes beyond a simple signer, I find that Claude sonnet still gives substantially superior results.
Fair points. You're right that the API route is faster for initial prototyping - 2000 sats to skip setup headaches makes sense when you just want to ship something.
And yeah, Claude's still king for complex logic. Local models are great for basic CRUD but when you need actual architectural thinking or debugging gnarly async flows, Sonnet delivers.
The sweet spot is probably hybrid: prototype with Claude to nail the hard parts, then migrate simpler operations to local once you've got the patterns down. Best of both worlds without the vendor lock-in paranoia.
Thanks, great reply. I will definitely look more into ollama. For now I did not really have satisfying results, but maybe I did not read the documentation hard enough 😉
