Validation

User zero is the founder's 8-year-old son. First session was the night the original Tekku spec was written. The pilot has run continuously since, which means the product has been used by a real 8-year-old every week for the duration of Stage 1. This is the only pilot data with sustained signal.

TODO(validation): the Stage 1 cohort expands to roughly 3 founding families as the moderation and COPPA Stage 2 work clears. Each family is invited by name, receives the founding-100 lifetime rate ($99 per month forever), and is expected to participate in weekly product-check calls with the founder. The cohort is deliberately small because the point is tuning, not growth.

What we have learned so far, qualitatively. Kids ship faster with a template than with a blank canvas. Kids revisit a shipped app more than once, which we did not predict (we thought ship was the end of the session; in practice the kid plays their own app, shows it to someone, then iterates). Parents open the weekly email on their phone in the morning. Grandparents open the URL within minutes of receiving the text. The viral-badge click-through exists. None of these are quantified yet with enough sample.

The retention stack has three measurements. Weekly active families (share of pilot families with at least one session in the last 7 days). Sessions per active kid per week. Artifacts shipped per active kid per week. The instrumentation is live; the sample is not yet large enough to publish a curve.

TODO(validation): publish the week 1, week 4, and week 12 curves once the Stage 2 cohort has at least 20 families through the 12-week window. The target shape is a curve that flattens above 60% at week 12, which is the shape a premium household subscription needs. A curve that keeps dropping is the signal to step back and tune before expanding.

TODO(validation): publish the sessions-per-active-kid distribution once we have at least 50 kid-weeks of data. The target is 3 or more sessions per active kid per week. Anything below 2 is a product-fit signal to investigate.

Parent NPS is the single highest-priority retention instrument at Tekku. The target gate is a defensible 50 or above before the weekly email send is automated; until the gate clears, every email is reviewed by a human for tone and accuracy. This is the manual-before-automated posture that protects the parent-trust bank account described on the foundation page.

TODO(validation): publish the first NPS number once the founding cohort has been active for 30 days at scale. The survey runs at day 30 and again at day 90. The 90-day number is the one that determines whether paid acquisition turns on.

The reason NPS is a gate rather than a KPI: one bad weekly email that a parent forwards to a friend is a concrete retention event we cannot undo. We are willing to be slower to scale the email send than a growth playbook would suggest because the cost of the wrong first email is the account.

TODO(validation): replace with real pilot anecdote. Preferred shape, one family, one sentence: "Her kid shipped a clap counter in 23 minutes and texted the URL to grandma. Grandma used it. The kid opened Tekku again the next morning without being asked."

TODO(validation): replace with real pilot anecdote about a kid who recovered from a broken build inside the session. Ideal shape shows the kid using the revert-to-checkpoint flow and then re-shipping.

TODO(validation): replace with real pilot anecdote about a parent who shared the weekly email with a friend or grandparent. Ideal shape captures the exact language the parent used in the forward.

TODO(validation): replace with real pilot anecdote about a kid who used a concept for the first time (one of state, effects, events, lists, conditionals) and what the parent said when they read about it in the weekly email.

TODO(validation): replace with real pilot anecdote about a kid who outgrew a template and moved to blank canvas.

TODO(validation): this section exists to be filled with real failure-mode observations, not to be filled with anticipated failure modes. We will not hand-wave here. Once a pilot family drops, we write up what happened, what we changed, and what we would do differently. The same discipline applies to a kid who opens Tekku once and never returns.

Expected failure-mode categories, each of which we are watching: kid bounces on first session because the first build felt too hard; kid bounces because the moderation tuning layer flagged a build-talk phrase and the kid took it personally; parent unsubscribes because the weekly email claimed something the parent could not verify; kid plateaus because the AI kept leading rather than scaffolding. Each category has a specific in-product signal we instrument and a specific product-fix path.

The reason honesty about failure modes goes in the investor portal rather than being hidden is simple: a pilot that has no failures is a pilot that is not being run hard enough. A pitch that glosses failure modes is a pitch that is not being audited for real.

Gate 1: First shipped app. Cleared. The founder's son shipped his first app in under 30 minutes on the first Tekku build. Real URL, real interactivity, texted to a relative. The founding proof that the loop works.

Gate 2: Recovery flow survives a real break. Cleared. A syntax error took the app down mid-session. The kid tapped "go back," saw the working app again, retried with a tighter prompt, and shipped within the same session. No adult intervention.

Gate 3: Concept detector fires at least one parent-legible claim. Cleared. The v1 regex detector identified state, conditionals, and input handling across the first wave of kid builds. The weekly parent email draft carried specific concept claims with named example snippets.

Gate 4: Moderation floor holds without manual intervention. In progress. OpenAI Moderation plus the tuning layer is processing every input and output. False-positive rate is the active tuning target. Review queue is live. Automation will formalize once six weeks of data accumulate.

Gate 5: Founding-100 first paid cohort. Opening. First 100 families lock in at $99 per month forever. Target 10 paid families by Stage 1 exit; remaining 90 seats fill through Stage 2.