'AI' and Csla #4685
Replies: 1 comment 8 replies
-
I mostly use Copilot as it is integrated with VS and vscode, and I have a license through work. I have also used Cursor quite a bit, and for "vibe coding" I think it is actually better, though copilot agent mode in vscode is catchingi up rapidly (imo). I also tried (just once so far) the GitHub feature where you can assign an Issue to Copilot and it'll try to fix the GH issue. That didn't work well at all for me, but I suspect it is because the Issue description was too cryptic, and perhaps the bug to be fixed was too complex (CSLA is not trivial after all!). I have used the GH ability to assign Copilot as a PR reviewer. That can be helpful in some cases. Outside of coding, I've spent countless hours over the past few days using ollama and OllamaSharp to run and interact with local LLM models (because that's free - no cloud tokens consumed). It has been an interesting learning experience, because we're still early enough in the wave that the docs are largely out of date and/or wrong, but there's enough tooling and GH repos that a stubborn person can get things working. It turns out that some of the smaller LLM models run fairly well on a machine with 32gb RAM and GPU. The bigger ones (bigger than 8 or 12 I've found) can barely run on my desktop, which has 64gb RAM and a gaming GPU. I think the limitation is the GPU only having 12 or 16gb of memory, so the models can't fit into the GPU, forcing them to run in the CPU instead. In any case, the machines generate heat like when I'm playing a high end game 🥵 |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
With all the buzz around AI tools for programming (productivity), I am curious about what fellow Csla developers use.
Thanks
Beta Was this translation helpful? Give feedback.
All reactions