Operating rules of the claude cadence
Run claude on a fixed weekday with a standing agenda. klitflempot recommends ninety minutes: thirty for new captures from interviews and demos, thirty for clustering and demotions, thirty for claim drafts and open questions. Protect the demotion ritual. Without it, libraries only grow and never get smarter.
Every promoted tile must answer: What mechanism repeats? Which roles confirm it? Which touchpoint showed residue? What claim would be false if a buyer contradicted us next week? If you cannot answer, keep the tile in quarantine.
claude also forbids silent math. Do not average Likert scores from eight interviews and publish them as “market sentiment.” Do not convert qualitative intensity into conversion forecasts. The cadence exists to improve language and focus — not to impersonate a survey vendor.
Artifacts klitflempot keeps in the claude kit
- Living committee maps per segment.
- Friction pattern board with anecdote vs. pattern columns.
- Language bank of buyer metaphors and polluted competitor phrases.
- Claim draft sheet with source citations.
- Open-question backlog that feeds next week’s interview probes.
Teams that skip observation often overfit to chatty champions. Teams that skip interviews often misread public artifacts. The claude cadence forces both streams into the same weekly room. That is the whole point.
What success looks like
Success is not a vanity dashboard. Success is a champion repeating your claim without coaching, a security review arriving earlier because you prepared for it, and a homepage line that survives legal because it never pretended to be a statistic. klitflempot measures the claude cadence by those operational tells.