A cohort whose six-week run opens and closes with a synchronous Source Pulse — a 20-minute Vibeclubs gathering where members share what they are building, not what they consumed — outperforms an otherwise-identical cohort that runs async-only on two outcomes:
- Six-week completion rate (members who ship the Week 6 distribution artefact): ≥15 percentage points higher.
- 90-day retention into Circle: ≥10 percentage points higher.
The mechanism we suspect: synchronous opening converts a roster of strangers into a recognisable cohort identity inside 20 minutes, and synchronous closing forces members to confront whether they shipped — which is the exact friction that drives the last 30% of work past the finish line.
This report is pre-registered. It is published 2026-04-22, six weeks before the first cohort starts. We will publish the result whether it confirms or falsifies the claim. Bookmark the open question on /research/generative-culture to track the resolution.
Pre-registered A/B-by-cohort design over three consecutive cohorts:
Cohort 1 — June 2026 (n = 50, capped) Treatment: synchronous Source Pulse
Cohort 2 — July 2026 (n = 40, capped) Control: async-only
Cohort 3 — August 2026 (n = 40, capped) Treatment: synchronous Source Pulse
We deliberately put the synchronous arm in the founding (June) cohort because the founding-cohort effect is large and confounding; we cannot cleanly test it. The clean comparison is July (control) vs August (treatment). June is included for descriptive purposes only and excluded from the primary inference. We acknowledge upfront that two cohorts (n ≈ 80) is small statistical power; this report is directional evidence pre-registered honestly, not a statistical proof. We would need ≥6 cohorts to claim a confident effect; we are publishing the design now so peers can run replications in parallel.
Cohort design is identical between treatment and control on every other dimension:
- Same six-week pillar arc (Identity → Catalogue → Library → Sessions → Governance → Distribution)
- Same Guardian-led async clinics (Veloura, Otome, Kaelith, Yumiko, Laeylinn, Sol)
- Same Frank cameo cadence (one live per week, 45 minutes)
- Same intake form, same capacity, same pricing, same Skool community shell
- Same Friday Wins thread, same Sunday digest
The only difference between control and treatment: presence or absence of the 20-minute synchronous Vibeclubs gathering on the opening Monday and closing Sunday.
Each Source Pulse session follows the same script:
Minutes 0-2 Open. Frank or a Stage-4 graduate frames the cohort question.
Minutes 2-12 Round-robin: every member speaks for ≤30 seconds.
Format: "I am building X. The pillar I want to nail is Y.
Here is the artefact I will ship by Sunday Wk6."
Minutes 12-18 Pair-up. Members get matched with one peer for accountability.
Minutes 18-20 Close. The Guardian present logs each commitment to the cohort vault.
Recording is opt-in. Members who decline still attend; their commitment is logged anonymously in the vault. Recordings are added to /community/cohorts/[slug] as an optional artefact for future cohorts.
The async-only control runs the same opening/closing prompts as a Skool thread instead of a live gathering. Members post their commitment in the thread and respond to two peers within 24 hours.
Best estimate before the data lands:
Outcome Control (async) Treatment (sync) Δ (predicted)
6-week completion ~55% ~72% +17 pp
90-day retention into Circle ~62% ~75% +13 pp
Member NPS at end of cohort ~7.5 ~8.4 +0.9
These priors come from two adjacent data points: (a) Frank's enterprise CoE engagements, where synchronous kickoffs lift program completion by 18-24 percentage points over slack-only kickoffs, and (b) creator-cohort published numbers from Maven and Wes Kao indicating ~15 pp uplift from any live ritual over pure async. Neither generalises perfectly to a 50-person solo-operator cohort; that is exactly why this report exists.
We commit upfront to calling this falsified if any of the following hold after the August cohort completes:
- Treatment 6-week completion is ≤ control + 5 pp
- Treatment 90-day retention is ≤ control + 3 pp
- Treatment NPS is lower than control on any subgroup of n ≥ 10
If falsified, the next report under this cluster will analyse why our prior was wrong and publish the corrected design. We will not retroactively change the hypothesis.
The narrow finding is about cohort design. The wider finding is about whether generative culture can be engineered at small scale. If 20 minutes of synchronous ritual moves a six-week completion rate by 17 percentage points, the implication for every solo-operator community — not just ours — is that the cost of "live" is over-estimated and the cost of "async-only" is under-estimated. Most operators kill live rituals first when time is short; this would suggest that is exactly backwards.
If the effect is small or null, the implication is the opposite: async-first is fine, and the operator energy now spent defending live rituals could go elsewhere.
Either way is useful. That is why we publish either way.
Resolution lands the week of 2026-09-21, after all three cohorts complete. We will update this post in place with the result, mark the open question on /research/generative-culture as resolved, and open the next question — likely a test of whether Stage-4 graduates teaching the cohort outperform Frank-led cohorts on member transformation outcomes.
Pre-registration timestamp: 2026-04-22. Method, predicted result, and falsification criteria are immutable from this point — any change requires a new report. Replication kit (cohort intake template, Source Pulse script, measurement instrument) published to the public Vaults of Benevolence under cohort-source-pulse-preregistration-2026-04. Built on SIP. If you run a parallel replication in another community, file a PR with your numbers and we will publish a meta-analysis.