← experiments

20 October 2026 · Built with Claude

Recall

A young dog goes after everything: balls, other dogs, squirrels, a dropped sandwich. Call it back. Every recall it answers builds trust, and every one it ignores costs some. Trust buys freedom: first the long line, then off the lead. Lose the trust and the lead goes back on.

A young dog on a short lead.

Trust 15%

Notes

Trust only moves when you call. A recall the dog answers raises it, most of all when it leaves something it wanted. One it ignores lowers it, because a cue that doesn't work teaches the dog to ignore the cue. Off the lead, an ignored call costs double, and the dog gets the sandwich, which it learns from too.

The long line is how trainers get round that: if the dog ignores you, reel it in. The recall still works, so the cue keeps its meaning, but the dog learns less from being reeled in than from choosing to come. Calling when it's easy, with nothing to chase, is cheap practice. Dog trainers will tell you it's most of the job.

It's the same question as handing autonomy to anything that learns: a lead you can pull while it's earning trust, more freedom as the record builds, and the lead back on when it slips. Your footsteps are still the beat, and the music is still the gap between you, busier and brighter when the dog's off after something.

The third dog walk, after Short lead, long line and To the park, and a cousin of Tandem in the loop. Built on 19 October for Day 20. All sound is synthesised as it plays.