[megaloblasto]

Awesome. Great info, thanks

Is this just a fun project for now, or could I actually benefit from it in terms of software production like I do with tools like claude code?

I am interested in carefully tailoring it to specific projects, integrating curated personal notes, external documentation, scientific papers, etc via RAG (this part I've already written), and carefully chosing the tools available to the agent. If I hand tailor the AI agents to each project, can I expect to get something perhaps similar to the performance boost of Claude code for $2000 (USD)?

If not $2000, then how much would I need? I'm pretty sure for something like $75000 I could do this with a large deep seek model locally, and certainly get something very close to claude code, right?

[megaloblasto]

I found the answer to this question here

https://www.youtube.com/watch?v=e-EG3B5Uj78&t=237s

"Run Deepseek R1 at Home on Hardware from $250 to $25,000: From Installation to Questions"

You can run Deepseek R1 with over 671B parameters at 4 T/s for ~$25,000 (USD). With AMD Ryzen™ Threadripper™ PRO 7995WX 96-Core, 192-Thread Processor and PNY NVIDIA RTX PRO 6000