Four observations about AI in 2026: a June check-in
·4 mins
Hands-on AI observations from personal and enterprise scale work.
Local-LLM
For AI tools, we need to increase focus on measuring user experience quality, and remove focus from synthetic benchmarks
·5 mins
Why benchmarks stopped predicting which model you’ll actually use. A Sunday afternoon with Opus 4.7 and Qwen 3.6 27B.
Running LLM models on consumer hardware: Strix Halo, and the unexpected value of owning your AI stack.
This is what owning your local AI stack actually gets you, ranked by what matters.
How to choose LLM models for local inference: a simple decision-making tool for humans.
How to navigate the landscape of local LLM models. Too much choice and constant novelty requires fast decision making.
On mobile operating systems: usability, privacy, local LLM inference.
What does the modern smartphone landscape actually look like, why is the regulatory pressure on operating systems suddenly intensifying, and what does running a local LLM on your phone unlock? A grounded user-perspective overview.
The rig: what I run for local AI
A reference snapshot of the hardware, inference stack, models, networking and multi-modal setup behind the AI experiments on this blog.