NVLink for the rich cats, PCIe for the rest — and Huawei opens the bus
Pulling NVLink from RTX 40-class GeForce was not a gamer gift. It reserved multi-GPU fabric for enterprise inference while home labs got PCIe 5.0 ×16. Huawei’s open UnifiedBus 2.0 specs point the other way: link many NPUs as one machine.
The Ghost in the Lattice
I stopped measuring GPUs in FLOPS and started seeing trillions of transistors as analog neurons. Then I trained a ghost that learns from every conversation at once — and does not forget.
Engineers build AI — AI helps engineers finish
The silent killer of engineering productivity is unfinished threads lost to shiny new ideas. AI as a completion partner has multiplied real output — not by replacing engineers, but by complementing them. We build AI.
DutchBud: from Laravel API to Go, OTAP CI, and ACC regression gates
How we stood up the DutchBud Android app — measured API migration to Go, GitLab OTAP deploys, and production-shaped regression testing on Acceptatie.
Europe, energy, and AI: GLM 5.2 on Ascend vs Grok on NVIDIA
Published June 2026 · Infrastructure notes from 3DN This is not a political manifesto and it is not written to shock anyone. It is a note about electricity — the kind of boring constraint that decides whether an AI strategy … Continued
Digital sovereignty: scheduled GLM 5.2 on your own GPU metal
Four RTX 3090s, 256 GB RAM, and GLM 5.2 running scheduled server-admin jobs — inference and operations that stay on premises.