Subscribe

A dangerously good use of your attention.

Put your email here. Future you will act smug about it.

Cybersecurity range operator turns a lockable network-isolation switch key beside glowing test machine status lamps.

The unrestricted AI chatbot — brakes, drivers and logs

Self-hosted AI infrastructure in an enterprise compute hall, with mesh-enclosed worker nodes, fiber trunks, and technicians replacing hardware.

Self-hosted AI agents — follow the bytes, not badges

Blue-hour shared-GPU inference room with open four-card server, engineers at terminals, patch leads, and cobalt window light.

When it breaks — pick your Ollama alternative by layer

Opened McDonald’s burger in branded wrapper on stainless remake shelf, harsh sunlight casting a long shadow.

McDonald's drive thru AI upgrade — will Archy work?

Bright local LLM testing studio with three inference stations: GPU tower, Apple Silicon machine, and background workstation.

The best local LLM in 2026 — three models, three jobs

Frankfurt data-centre district with layered security fences, cooling structures, closed gates, cyclist, and distant commuters in winter sunlight.

No exit plan, no data sovereignty — even with EU hosting

Used RTX 3090 graphics cards running workload tests in open rigs beside power supplies and cooling ducts.

Before you buy an RTX 3090 — run your workload first

RTX 3090 graphics card in anti-static tray at reseller counter, showing cooler, power connectors, PCIe edge, and 24GB VRAM.

Why 24GB Still Matters — RTX 3090 for Local AI in 2026

Formula 1 garage with matte car, lowered front-wing elements and open rear-wing gap during static DRS check.

DRS F1 explained — why the wing no longer wins passes

Incident-reconstruction suite viewed through observation glass, investigators examine AI agent authorization logs, benchmark stations, and network taps.

At Hugging Face — AI agent authorization had no veto

A character AI alternative — can you take it with you?