Weeknotes 17 April, 2026

A lot of this week was about measurement. Not in a metrics dashboard way, but in the quieter sense of pausing to ask whether the way something is being done is actually the right way to do it.

The most energising part was starting to explore how we might train Claude to replace a custom GPT we use for content work. It’s early, and there’s a lot to figure out, but there’s something satisfying about moving from “this tool exists” to “this tool could be shaped.” The Claude Code 101 course helped sharpen that thinking – less about what AI can do in the abstract, more about what I can actually do with it in practice.

I also spent time measuring how long a manual content review takes compared to doing the same work with Claude or Cursor. That kind of comparison is only useful if you’re honest about what you’re actually measuring. Speed is one thing, but quality and consistency matter too. Still, it’s useful to have the numbers. It makes the case clearer, whichever direction it points.

The content audit review was a different kind of work. A colleague had proposed archiving a number of pages. This was without explaining why those pages, or whether the content was adequately covered elsewhere. Giving useful feedback meant thinking carefully about what I actually needed from my colleague. Not just a decision, but the reasoning behind it, so someone else can follow the logic and trust the outcome.

Somewhere in the middle of all this, I wrote my own job description. I’ve been working without a formal one for a while. Writing it down – accurately, without either underselling or inflating – turned out to be more clarifying than I expected. The next step is to write what the role could look like at the next level, as part of the work I’m doing towards a promotion. That felt like the right order. Understand where you are before you describe where you’re going.