Wicklee Blog
- Apple Silicon Thermal Throttling: What Other Monitors Miss
Your M2 Max is throttling and you don't know it. Activity Monitor shows green. Wicklee reads the same IOKit classes Apple itself uses — here's how, and why it matters for local AI inference.
- Hardware-Aware Observability for Self-Hosted AI
Generic observability stops at 'GPU at 80%.' Self-hosted AI needs to see deeper — SoC power, thermal penalties, per-model VRAM, inference state. Here's what hardware-aware means and why it matters.
- Runtime Config Surface: See What Every Node Is Actually Running
Two nodes loaded with the same model produce different WES scores. Why? Wicklee v0.9.0 surfaces the launch config of every Ollama, vLLM, and llama.cpp node so you can find out in one click.
- WES: The MPG for Local AI Inference
tok/s tells you how fast. Watts tells you how hungry. WES tells you if the tradeoff is actually worth it.