Field notes

Rob Burke

Latest side quest: Oct 2026

AI agents, side quests, and 25 years of synthetic characters

A swimmer in an ESWIM cap mid-breaststroke, lane ropes blurred behind.

Side Quests · Feb 2026 · No. 07

Swim Meet Photographer’s Assistant

Why not ALSO use Claude Code + GitHub to collaborate with an AI when you’re the Swim Club’s Meet photographer?

[Jun 2026 edit: this was the spark of something that has evolved a long way, including a Character-based AI workflow inspired by a Group of Seven painter, and I now use my Obsidian notes rather than GitHub as a collaboration surface,.. but still do publish to a private GH repo on event day because the format is so accessible in a browser on a pool deck where the network signal is a hot mess.. next writeup is here]

Could be a case of “hammer, nail.” But.. it felt obvious that the best way to plan and execute being ESWIM Club Photographer for a meet, optimizing for a specific set of goals, was to give Claude a bunch of context, activate Planning Mode and brainstorm. Then use Markdown on GitHub as a collaborative “surface” to optimize the shoot in realtime while running the pool deck.

My input context to Claude: “heat sheets” showing who’s swimming when and in which lanes, which side of the pool is easiest to navigate, some ‘banger’ shots I wanted to try for, etc.

The key outcome was a chronological “game day cheat sheet” on my phone. Benefits:

  • GitHub repo allowed iteration and refining from my phone using a “remote control” Cloud instance of Claude, on deck with SLR in other hand
  • Markdown file acted as “living document” in my phone browser
  • Claude fearlessly dissected swim-meet-domain-specific file formats to extract information I wanted (swimming time standards in a weird binary format)
  • All we learned is memorialized and ready for future swim meets

All of that was essentially pages from my vibe coding playbook, applied to a different domain entirely.

There are some lessons learned in here that I suspect will generalize for me:

Day 1 (get new gear): Rent fancy lens, fumble around, go home frustrated and cook up this plan while commiserating with Claude

Day 2 (completionist, too much AI): Execute a ton of technical shots, frantically try to “catch ’em all” and largely succeed, feel great in the moment ….. but get frustrated again while editing in LightRoom, feeling like I missed all the personal moments. Refine the .md file to add the emotional moments I wanted to catch, and the WHY BOTHER to complement the WHAT.

Day 3 (storyteller): Remember it’s all about the story, become AI-empowered narrator, end satisfied.

I know a lot of folks won’t want to use this set of tooling and that’s perfectly fine. But it worked incredibly for me in Feb 2026.

Pics or it obviously didn’t happen! (shared humbly – this is the start of my swim photographer journey; I’m still working on it, folks.)

[a quick postscript: this turned into a huge source of inspiration for me, on the AI front and otherwise. More pics here.]

This later grew into Brand New Renaissance, where I taught an AI my own editing taste.

Backstroke swimmers arch backward off the wall at the start of a race.