Not true, for $2500 I have produced a shit ton of code and applications. This was created for around 10% of a monthly 200 USD Claude subscription.
If you had $2000-$2500, what would you build or buy?
- Thread starter Greg_E
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
That's pretty cool app.
its funny, one of my ideas for testing app building with my local llm was a home based chat app that runs locally.
i never spent a ton of time on it but it did basic functions of chat messages stored in a db. I didnt code any of it. just helped the llm along the way.
I mostly need to dial in my llm models/setup with these test apps.
Now that I am mostly dialed in, I am now using my setup to code a websphere ND app that helps set settings on cluster/servers and backup profiles of Dmgr and nodes etc.
Plan is to have it assist with new environment build and automate the small details settings that need to happen.
Again slow going but its free (well minus the hardware costs).
Chat applications are actually super simple to create, because the LLM already knows the specification. I just said, do whatever Slack or Mattermost do. But keep the backend in Spring Boot, Keycloak, PostgreSQL and Lucene. A familiar and simple stack is easy to review and maintain too. And if you want to integrate it with enterprise (Entra ID), thats fully supported with Keycloak.
That's a cool project, sounds like you could run it on openSUSE LEAP Micro and cut the foot print by a small amount. I'm going to try and remember to grab this once I get my lab back up, will be fun to test and see what I can use it to do around my production system.
Did you use Claude paid or one of the local Claude models to make this?
Finally, I'm in the USA, and I'm loving this Euro-centric development phase... It really just means better "stuff" for those of us who want to be outside the Microslop box, and really I don't have much love for IBM/Redhat these days.
If that is without a GPU I would build a dual CPU Epyc gen 2 or 3. 2x 64 core CPU, DDR4 ECC Reg.
I did some learning with 3 x P100, thats 48GB VRAM with EPYC 16 core 512GB system ram. Could ran Qwen 3.8 27B Q4 ~30toks. P100 are cheap ($80) and not well supported, but getting those out of case costs same as GPUs. Planned to have 4x or more, but one died. Got modded 170 HX 64GB for $1.6k and it does ~55toks on same Qwen model that is quite good. Much easier to get support as Ampere is still 'recent'.
Well, that's probably for the best - it's hard to justify the hardware and energy usage for LLMs in 2026 if you don't have a secondary use for it.
Yes, but the problem is that with todays prices you don't get much useful hardware at all for that budget, which sucks.