Pet · Product · 0 → 1 · 2026

Lumos — bedtime stories

Built an end-to-end clickable prototype of the whole funnel (landing → personalisation → generation → payment → narrated series) and a calibrated prompt engine. Preparing to launch.

TL;DR

An end-to-end prototype of a bedtime ritual story: from landing to a narrated series — on mocks, no LLM API

  • 01Designed the product from Discovery and AJTBD segments to RAT and a funnel with built-in analytics
  • 02Built a clickable end-to-end prototype (vanilla JS, no build): landing → form → generation → free episode → mock YooKassa payment → series of 5 + upsell
  • 03Built and calibrated the generation prompt engine (~5 min of voice per episode) and a reference demo series of 5 episodes

The idea

Children aged 3–7 are often afraid of the dark, and bedtime turns into a struggle. Lumos is a personalised 1–2 week ritual story series where the hero resembles the child and, step by step, overcomes the fear. Not a one-off story but a nightly ritual. The personal goal — to test the skill of solo-launching a digital product and put it in the portfolio.

What I did

Went through the product cycle: Discovery and market, segments and jobs (AJTBD), risky assumptions (RAT), a PRD, a “touch → purchase” funnel with built-in analytics. Built a clickable end-to-end prototype in vanilla JS with no build and no external APIs: landing → personalisation form (name, age, gender) → mock generation → free first episode → mock YooKassa payment → a series of 5 episodes with an upsell. With real narration (Web Speech / server TTS, voice choice) and an on-screen funnel event panel. Separately built and calibrated the generation prompt engine (~550 words ≈ 5 minutes of voice per episode) and generated a reference demo series of 5 episodes with a coherent dramatic arc.

Challenges

Make the product convincing and testable without wiring up an LLM API or real payments — all generation and payment on mocks and heuristics, but in production form. Self-employed compliance: no medical promises — wording like “helps gently process and fall asleep calmly,” not “cures insomnia.” Privacy: personalisation without uploading the child’s photo. Distribution with no budget — betting on organic content.

Insights

The key pivot: the user’s job isn’t “one story” but a 1–2 week ritual series. That shifted the whole product from a one-off purchase to recurring value. For fast validation I took the sharpest segment (the “sleep rescuer”). And I chose not to run isolated tests but to build the MVP around the funnel with analytics from the start — the data will gather in production. The build-in-public format also works for the portfolio.

Next

Wire up generation via the Claude API and real YooKassa, ship to production, launch an organic content plan, and gather the first legal sales.