Building a Private AI OS on 4x RTX 2080 Ti
askincihan · reddit · 2026-09-01
The author built a private AI operating system called OMNI CORE on 4x RTX 2080 Ti (22GB) GPUs. Users interact solely with a main agent (OMNI MAIN), which decides whether to answer directly, browse, use the terminal, inspect files, work with GitHub, retrieve memory, or delegate to specialists like CODER or FLUX.
Stack & Hardware:
- Hardware: 4x RTX 2080 Ti 22GB (Total 88GB VRAM), 129GB RAM.
- Models: Main (Qwen3.6-abliterated 27B), Coder (Qwen3-Coder 30B), Image (FLUX.2-klein-9B).
Key Features:
- Resource Manager/Scheduler: Defines GPU topology and priorities (MAIN is highest). Tracks job and resource lifecycles (READY/QUEUED/RUNNING). Supports leases, heartbeats, and stale recovery, ensuring that process exists != valid job.
- Governance & Acceptance System: To prevent unreliable "agent writes code -> done" workflows, it introduces a canonical project state (ARCHITECTURE, ROADMAP, STATE, ACCEPTANCE, LEGACY). All substantial changes must follow a strict process from definition to bounded implementation.
More from coding & agent
- MongoDB CTO on Database Architecture Evolution and the Unsolved Problem of Agent Memory — The Cognitive Revolution · 2026-09-01
- Langfuse Releases Evaluator Template Gallery for LLM Apps — colinmcnamara · 2026-09-01
- Super Library Agent: Joint Generation and Maintenance of Multiple Applications — kaist-ai · 2026-09-01
- DAX-1 open weights released, outperforming Fable 5 on spreadsheet benchmarks — saheedniyi_02 · 2026-09-01
- AI develops 27B model autonomously, Recursive Self-Improvement achieves industrial results — 机器之心 · 2026-09-01
- Public Safety AI: How Peregrine Uses Agents to Solve Cold Cases — Training Data (Sequoia) · 2026-09-01