Insights & Engineering Notes

The Genworx Blog

Practical perspectives on Zero Trust networking, managed IT on owned infrastructure, and private AI platforms.

LATEST · 21 OCTOBER 2026

DeepSeek V4.1 Flash: 1M Context on 4 RTX PRO 6000s

Four PCIe Blackwell cards, no NVLink: 1M-token context at 149 s cold prefill and 212 tok/s decode, now with three concurrent sessions.

Read the article →

10 July 2026

Local AI Inference: Taking Back Control of Your Data

Cloud AI APIs are convenient — until you consider data sovereignty, latency, and unpredictable costs. Local inference changes the equation.

Read article →

7 July 2026

Four Mac Studios: A 4x M3 Ultra MLX Cluster

Four Mac Studios, 1 TB of unified memory and a Thunderbolt 5 mesh, running one model between them — on an inference engine we wrote.

Read article →

1 July 2026

Why Zero Trust Is Replacing the Traditional VPN

Traditional VPNs grant broad network access once you're in. Zero Trust flips the model: identity-based, least-privilege access to individual services.

Read article →