AI & Building: Local LLMs, Agent Workflows, and Real Apps

I write about AI the way I actually use it — not as a hype cycle to react to, but as a set of tools I run daily to build things, verify things, and get more done with the same 24 hours. Most of what's here comes out of a home lab where I run a multi-machine local LLM fleet and a growing set of agent workflows, plus the small apps I've shipped because I wanted them to exist.

Recurring threads: standing up local inference on consumer hardware instead of renting cloud APIs, wiring agents and cron jobs into an actual operations layer instead of a chat window, and treating a single AI-drafted output as a first draft that needs a second, often adversarial, model to catch what it got wrong. I also write about the unglamorous stuff — SEO cleanup, GA4/BigQuery pipelines, funnel apps — that AI made faster without making it magic.

If you're evaluating whether local models are worth the hardware, trying to get agents past demo-ware into something that runs unattended, or just want to see what a working AI-augmented workflow looks like from someone building it in public, this is the category for you.