Benoit C. - local AI & security
Security, local AI and sovereignty.
I publish here what I learn while building AgentVUE: security, AI that runs on your own hardware, and the day-to-day of software driven by AI agents that work the way employees would.
These notes are personal. The AgentVUE solution and its documentation live on agentvue.ai.
Featured
All articles
What is the origin story of AgentVUE?
The personal story of how AgentVUE began: origins, systemic risk and a philosophy of vigilance.
Latest articles
40× RTX 5070 Ti: the rig behind AgentVUE
640 GB VRAM · 5.5k+ t/s (PP) · 800 t/s (8 agents)
The final monster: 32 GPUs and 1 TB of VRAM
1 TB VRAM · 9.7 t/s (Kimi K2.6) · 4,800 W peak
Scaling to 16 GPUs: DeepSeek v3.2 and Qwen3.5 at home
512 GB VRAM · 32 t/s (Qwen3.5 397B) · ~€5.6k
My first local AI rig: 8 GPUs, 256 GB of VRAM, $880
256 GB VRAM · 26.8 t/s (MiniMax-M2.1) · $880
The journey at a glance: 8 → 16 → 32× MI50 → 40× RTX 5070 Ti - with all the details in the articles.
About
More than ten years in data, security and identity - Safran, Idemia, EDF - then founding AgentVUE.
