Blog
Field notes
Notes on AI hardware, low-power compute and systems architecture.
October 31, 2026
Self-Hosting Security Risks: Built and Abandoned Systems
Self-hosting is booming. But systems freeze the day they are built: 185 kernel releases missed, 763 CVEs in one week, a 74-day patch lag. Where is yours?
self-hosting · security · patch management · linux kernel · homelab · infrastructure · privacy
October 24, 2026
RTX 5080 24GB and the 16GB Tier: A 2027 Price Forecast
Nvidia has not discontinued the RTX 5090. The silicon and memory modules behind it are moving to AI, and that is what prices every card below the flagship.
hardware · AI hardware · GPUs · GPU pricing · local AI · capacity planning
October 15, 2026
The Next CPU Shortage: Why Your AI Agent Needs a Plan, Not a Faster GPU
GPU and memory shortages were only the first waves. The next one is about the most ordinary chip in the box — and the fix is not a bigger budget, it is knowing your hardware and sizing your work to fit.
hardware · AI hardware · AI agents · infrastructure · capacity planning
October 12, 2026
The Physics Behind the Screen
A demo on a screen, a custom chip running somewhere unseen — and the moment the audience understood that software always runs on something physical. Why we can't imagine the hardware we use, told without jargon.
hardware · AI hardware · how computers work · systems
October 11, 2026
Low-Power AI Accelerators and FPGAs: Inside Apex Compute's Unified Engine
Joules per token, not TOPS. A close look at Apex Compute's FPGA-based transformer accelerator — real benchmarks against NVIDIA Jetson Orin Nano, and what a 12nm tape-out changes.
edge-ai · fpga · ai-accelerator · low-power · benchmarks
- AI hardware
- Low-power compute
- Local LLM
- Systems architecture