LOCAL AI BOX

A local AI workstation replacing $700/mo in cloud plans with a used GPU and open models.

Open source

Description

A custom-built local AI server setup utilizing a $500 used NVIDIA RTX 3090 to host a suite of open-weight models via Ollama. It features a multi-layered agentic architecture where 'Jev' acts as a decision layer for data relevance and 'Kimi K3' handles writing tasks, all accessible remotely via Tailscale. The system successfully replaces professional tiers of Claude, ChatGPT, and Grok for the cost of electricity.

Descriptions, tags, and model credits may be AI-generated or inferred from public sources and can be incomplete or wrong. Learn more.

Details

Why this is interesting: A sophisticated example of multi-model routing that reduces agent costs by 90% by separating decision logic from…
Attributed to: @starmexxx
Built with: Jev Kimi K3 Qwen3-Coder gpt-oss-20b
Links: View tweet
Tags: #local-ai #agent #rtx3090 #self-hosted

Discovered: X
First posted: X · Oct 8, 2026
Listed: Added 7d ago
Build: Moderate
Models: Strong

LOCAL AI BOX

Description

How it grew

LOCAL AI BOX

Description

How it grew