The measured run
One run, published whole — including the parts that did not work.
We have no customers and therefore no testimonials, and we are not going to invent any. What we have is this: one end-to-end run on our own production system, recorded to a receipt, and published here without editing. Every number below is read out of that receipt.
- Recorded
- 2026-08-17T13:51:09.577Z
- Account
- created during the run · solo
- The one sentence typed in
- “AI-coached puppy training subscription for busy first-time dog owners”
The running count.
Zero. No customer has completed a build yet — a build counts only when all 8 gates of a customer’s journey are accepted, and none has been. This number is read from the live journey ledger, at most a minute stale, so when the first one lands it will move on its own. The measured run below is our own and is deliberately not counted.
Counting real customer builds only. Builds by our own accounts, the platform’s seeded demo tenants, sandbox and probe tenants, and synthetic test personas are excluded — and so is any build we cannot attribute to a real owner, because a count that cannot vouch for its rows is not a count.
Two panels came back empty.
This is first on the page rather than last. In the run below, two of the nine artifacts produced nothing — and this is why competitor analysis is not on the list of what the product gets you, on this page or anywhere else. It is a gap. We are not going to sell it as a feature until it fills. Here is what the run actually printed where the content should have been:
- Competitors — the artifact reads
“No competitor scan yet” - Positioning — the artifact reads
“competitor gap × the avatar's top pains. Regenerate to fill.”
Verbatim from docs/_demos/runs/run-2026-08-17T13-44-19-657Z/a01-solo-petcoach/artifacts/MARKET_INTELLIGENCE.txt.
Every gate, every timestamp.
Seconds are measured from the moment the idea was submitted. “Generated” is when the artifact came back; “accepted” is when it was read and accepted.
| Gate | Generated | Accepted | Status | Artifact |
|---|---|---|---|---|
| Validate idea (score + SWOT) · scored 82 | 9.2s | 11.2s | ok | 2123 B · md5 9f2776908a52 |
| Market Intelligence | 33.5s | 65.5s | ok | 4044 B · md5 dfd8b6e27a81 |
| Plan Studio | 91.9s | 124s | ok | 5245 B · md5 dc9286a10721 |
| Brand Studio | 134.4s | 166.4s | ok | 2353 B · md5 fdcabf6fe862 |
| Message Studio | 188.9s | 221s | ok | 2843 B · md5 14f908e9cd55 |
| Offers Engine | 232s | 264s | ok | 1429 B · md5 b74fbd4f3de0 |
| Funnel builder (WYSIWYG) | 286.5s | 318.4s | ok | 2133 B · md5 dff22a49c865 |
| Campaign Planner · scored 85 | 330.1s | 373.4s | ok | 2849 B · md5 5a36f8ba6e28 |
| Launch business (capstone) | — | 390.5s | ok | 1557 B · md5 39b2285e88a9 |
What was standing at the end.
- Funnel published: yes · the published URL answered HTTP 200 · architecture chosen: Subscription
- Published-page quality check: 100/100
- Buyer journey run against the published surface: PASS
- Adversarial buyer probes: adversary: 8/10 defended; 2 blocked
The published URL is not linked here on purpose: it was served on the run’s own sandbox host, so linking it would imply a live page you could visit, which is a bigger claim than the receipt supports. What the receipt supports is that it published and answered.
Every finding — 10 of them.
These are the run’s own findings, blocked legs and unset-up probes, printed in full. A blocked test is a gap in the test, not a pass — we count it here as a gap and never in the 8/10 above.
- ledger-taint
self-signed payment event(s) — this run signed its own checkout.session.completed, so the webhook leg is asserted, not observed; the ledger demotes it to PARTIAL(missing-leg: real Stripe event), never PROVEN: evt_selfsigned_cs_test_b1DXK38PEKWjN8djOTR1Q9lE1pBc6CY31wpFPY5m2WXnnmUmMwvlwU5ajM
- Buyer leg B1 — undefined
- Buyer leg B3 — undefined
- Buyer leg B4 — blocked
no delivery token from the opt-in submit for this buyer/archetype — a FILE archetype delivers via the PURCHASE (checkout → /d), proven in the money receipt (deliveryTokenId + /d 200) and B3, not the opt-in; an ACCESS/booking archetype delivers via the portal/booking. The /d route IS landed and proven (#858/#899). Blocked = honest non-coverage for this leg, not an unbuilt route.
- Buyer leg B5b — blocked
no email_log/outbox read seam — the reply→buyer notification is not HTTP-observable (depends R3-E1-2/E7-10)
- Buyer leg B8 — blocked
no email_log read seam — the dunning send is not HTTP-observable (depends R3-E7-10 read seam)
- Buyer leg B10 — blocked
win-back lapse is wall-clock-gated: buyer-lapse-sweep only emits buyer.inactive past the day-threshold and the harness has no test-clock. Honest block (named gap), not a fake pass nor a fabricated lapse.
- Buyer leg B9 — blocked
unsubscribe token is not HTTP-returned by opt-in and no buyerId resolved to erase (buyer-identity read seam absent) — B9 blocked, ordered LAST (destructive)
- Adversarial probe "delivery-token-reuse" — blocked, not defended
single-use /d delivery token reused after consumption · harness result: no delivery token minted
- Adversarial probe "booking-flood" — blocked, not defended
one email books every open slot (no per-email booking limit) · harness result: no booking surface published