The Learning Loop is the Atom

The smallest unit of work in LDD is a complete learning loop:

1. Frame a hypothesis (what do we think will matter?)
2. Slice it into the thinnest possible increment (thin enough to validate quickly)
3. Build it (with or without agent assistance)
4. Validate the assumption (did customers care? did it work?)
5. Decide what to do next (double down, pivot, or abandon)

This loop should complete in hours to days, not weeks. The speed isn't about rushing. It's about closing the feedback gap between assumption and evidence.