Kelvin Zhang
Co-founder of Gleam AI with Charles Liu; self-described shy kid turned designer of the app’s Duolingo-style experience. Drives mood boards, design systems, onboarding and paywall setup (hard paywall, 3-step paywall, yearly default).
Strong on: structured design process for AI-built apps, day-zero onboarding, paywall structure.
Bias to keep in mind: one app, self-reported numbers; speaker labels in the transcript are inferred.
Sources: sources/youtube/app-leaders-by-adapty/2026-10-15-how-two-ex-engineers-built-a-140kmonth-social-skills-app-gle.md.
Notes citing Kelvin Zhang
- Content first, app second — launching from a personal brandStella's founder (1.2M Instagram / 2.8M TikTok) built the audience first, mined comments for the product gap and reports >$300K MRR within ~2 months; she argues virality is a learnable skill (new accounts to 40K–700K followers) — but Symmetry's YouTube community did not carry its app.
- Copy or innovate? Differentiating in a "saturated" app categoryJake Castillo's view that saturation means proven demand (win with a novel, easier feature or a hyper-specific niche, and avoid social apps), Gleam's "copy Duolingo's defaults, add a twist", and Nami/Storemaven's warning against copying competitors' pricing and paywalls without knowing why.
- Deciding what to build — OKRs, funnel analytics and A/B-testing every product changeSymmetry's (~$160K/month) product process — quarterly OKRs, event tracking on every onboarding screen, user interviews for the "why", problem-first hypotheses with guardrail metrics, ≥2-week A/B tests in PostHog — where a one-screen commitment step lifted retention 3% and a redesigned home screen moved nothing.
- Finding the one repeatable organic format for your appHow Gleam (non-visual social-skills app) spent 3–4 months testing copied formats before long open-loop storytelling videos worked, why its 100K-follower POV accounts barely converted, Stella's savable "repeat for 7 days" affirmations and ManyChat comment-to-DM, Symmetry's lock-screen-notification format (17M views) and sub-format tracking — then recreating the winner "to infinity".
- How often to run paywall tests — and why a 5% ARPU lift can be your whole ad marginCal AI ran ~4 paywall tests a week (~300 in 18 months), single tests worth $10–50K/month, starting from ~100 users/day; Gleam argues a 5–6% ARPU lift can be the entire margin on paid ads and unlocks channels (and acquirers); treat the paywall as a never-finished marketing asset.
- Personalising the paywall with what onboarding (or the ad) already told youCarry intent signals into the paywall — Fastic's 20–30-step onboarding (~80% completion) puts the user's target weight on the paywall; Gleam returns a persona scorecard and course picks before asking for money; Apple Ads keyword or ad group can pick the paywall variant.
- Setting your first subscription priceStart higher than feels comfortable — Ravi Mehta's $20–30/month "high waterline" doubles as a PMF test — then anchor on competitors, cost-plus-perceived-value (AI apps), ROAS/payback math and the value of the user's whole problem; text-AI wrappers cluster at $30–40/year, niche custom-prompt chat apps get $20+/month.
- Stuck with no traction? Remove the friction before the first successVisualizee sat at ~$100–150/month for two years until a node-graph UI became a chat with a prompt assistant (and one-time payments became subscriptions) — ~$10K/month; Cal AI won by turning calorie "homework" into one photo; activation friction defines what a category can do.
- The three-screen pre-paywall sequence (explain the free trial)A 2–3 screen sequence before the paywall ("try free", "we'll remind you before it ends", trial timeline) that raised trial starts 56% in one app; Blinkist, Rest of Me, Cal AI variants.
- UGC, influencers or paid ads first? Choosing by cash, stage and marketWhat each channel buys (UGC = a swing at the algorithm, influencers = borrowed trust, paid = clicks, so make it hard-sell), Jake Castillo's rule — no money → UGC/self-made content, some budget → influencers, cash-flow not a concern → paid — and how Gleam, Symmetry and Cal Buddy actually sequenced channels.
- Weekly vs monthly vs yearly plans — and which plan gets the trialMonthly plans usually underperform; test weekly vs yearly per category, and put the free trial on the plan with the highest LTV — moving it from yearly to a 3-day weekly tripled sales in one app.
- What breakout indie app case studies have in common (Cal AI, Gleam, Stella, Cal Buddy, Symmetry, Visualizee)Side-by-side numbers from six 2024–26 founder stories ($10K to $40M ARR) and the patterns they share — a big proven category, one low-friction core action, one repeatable distribution engine found before scaling, hard paywall with yearly/trial, relentless testing — plus where they contradict each other.