A Skill is More than Markdown
In which I work through the actual lifecycle of an official company skill, and find at least six roles that don't exist in most org charts yet.
Hey, I'm Dachary Carey
Tools, apps, and strong opinions about coffee
Can Agent Skills Make Output Worse?
In which I discover that a skill made LLM output 66% worse, and dig into why authoring official skills is a full product lifecycle, not a deliverable.
Astro Removed its llms.txt
In which Astro removes its llms.txt entirely, and I explore the result.
GitHub Reimagined llms.txt as an API
In which GitHub replaces the standard llms.txt URL list with a purpose-built agent API.
What an Agent Score Can Tell You
In which I investigate what agent scoring tools actually measure.
Designing an Agent Reading Test
In which I try to give people tools to understand how agents read web content, and where they fail.
Measure Agent Web Traffic Redux
In which I revisit measuring agent web traffic, and dive deeper down the rabbit hole.
Confident-Sounding Gibberish
In which I share research about why LLM-generated output is hard to fact check.
The Verification Gap in AI Content Pipelines
In which verification is the hardest unsolved problem in AI content pipelines, and most organizations don't know it.
How to Evaluate a Platform-Written Spec
In which I find that a platform-published spec's omissions track financial incentives.
