Why I started AI Learning Press

I've been building software for a little over twelve years. For most of that time the hard problems were the usual ones: state, data, people, deadlines. In the last couple of years a new problem slid in next to them, and it's the one I keep coming back to.

The model almost never has the context it needs.

Not "the prompt is bad." Not "we picked the wrong model." The system just didn't hand the model the information a competent human would have had in front of them — the ticket history, the schema, the three Slack threads where the decision was actually made.

This site is where I'm going to write about that gap and the work that closes it.

What I'll write about

• Context engineering. How to find, rank, trim, and format the right information so a model can actually use it. Retrieval that doesn't hallucinate relevance.
• Evaluation. How to know if a change made things better, without asking three people to "vibe check" it.
• Automation inside organizations. The process design that makes an AI workflow trustworthy enough that people stop double-checking it.
• ML and mapping side projects. Python, geospatial data, and the occasional detour into something that has no business value at all.

How this site works

It's a small Next.js + MDX site. Posts are markdown files in a content/ folder. Each one carries a short, hook-first "LinkedIn version" in its frontmatter so I can share consistently without rewriting the same thing twice — you'll see it at the bottom of every post.

title: "Why I started AI Learning Press"
description: "A place to write down what I'm learning…"
tags: [meta, ai-engineering]
linkedin: |
  The hook goes here. First ~200 characters matter most.
hashtags: [AI, ContextEngineering]

If any of this is useful to you, the best way to tell me is on LinkedIn. More soon.