Practice
Scaling LDD
Growing from one loop to many, solo or as a team.
For solo builders
Once you're comfortable with single loops, the temptation is to run many in parallel. Be careful. The bottleneck is your attention, not your build speed.
- Sequence over parallelism: Run one or two loops at a time. Finish validating before starting the next build. Your AI agent can build fast; you can only interpret evidence at human speed.
- Keep a hypothesis backlog: Write down the next three things you want to learn. Resist the urge to build all of them at once.
- Weekly reflection: Set aside 30 minutes a week to review: What did I learn? What did I ship that I never validated? What should I kill?
For teams
- Dual-track: Separate discovery (framing next loops) from delivery (validating current loops). The key insight: this isn't "PMs do discovery, engineers do delivery." Anyone can contribute to either track. The split is about focus, not roles.
- Loop portfolios: Run multiple learning loops in parallel at different stages of maturity.
- Learning cadence: Weekly or bi-weekly "decide" sessions where the whole team reviews validation results and frames new hypotheses together. This is where the PM-engineer-designer distinction matters least. Everyone is looking at the same evidence.
- Continuous shipping: As loops validate, ship incrementally. Don't batch everything into a release.