Weeknotes for the week finishing Friday, 8th May 2026.

Two weeks’ worth of weeknotes this week.

Last week I was in Birmingham (“You feel as if you’ve been projected into the 21st Century”), meeting with my Bennett Institute colleagues. There were sessions on LLMs, OpenSAFELY, and OpenPrescribing, as well as time for five-a-side football, lawn games, and crafts.

LLMs

The sessions on LLMs were useful and interesting, but I don’t think we’ve come to terms with how they’re changing our profession. Peter showed us how he’d use a coding agent to ship a feature — ask it to make a plan, review the plan, ask it to execute the plan, test the feature, review the code, ask it to create a logical sequence of commits, create a PR — and we discussed how we’d review the PR. We also discussed whether we’d review the PR. We didn’t reach any firm conclusions, sadly.

Two weeks ago, I asked “What’s the point of code review?” I’m still unsure, but I don’t think “We shouldn’t review AI-assisted PRs”, “We should review them, just as we would review any PR”, and “We should ask an LLM to review them” are satisfactory answers.

Patterns

This week, between meetings, I’ve been thinking about Architecture Patterns with Python by Harry Percival and Bob Gregory. I really like the book, especially the distinction between data models and domain models, and how we can make the distinction clear by following the dependency inversion principle.

The code examples draw on SQLAlchemy and Flask, and in the six years since the book was published both libraries have changed a lot. So, I’ve spent some time thinking about what the code examples would look like now, in 2026.

Metaphors we live by1

I’m not the first to have noticed that the debt metaphor is everywhere. Alongside technical debt, we now have cognitive debt, intent debt, and operational debt. Cognitive debt is, I think, the same as comprehension debt. And, switching metaphors, all seem to relate to cognitive offloading and cognitive surrender. Keeping up with AI is hard; keeping up with the metaphors is even harder.

  1. I’ve not read the book. Looks interesting, though! ↩