How a 17-year-old built a $1M/month AI calorie-tracking app

Executive overview

Most calorie-tracking apps are bloated and outdated. Cal AI replaced them with a single action: photograph food, get macros instantly.

Zach Yadegari and two high-school co-founders launched in May 2024 and hit $1.1M monthly revenue by November — bootstrapped, 50% margins, 12-person team.

The engine is a disciplined influencer marketing system built around CPM/RPM economics, not brand deals or paid ads.

The core insight: pick an idea that is inherently viral, nail the distribution before you build, and hard-paywall from day one.

From gaming website to app studio

• Started coding at 7; first game on the App Store at 12.
• At 13 built TottallyScience.co — an unblocked games site for students — grew to 5M users via personal TikTok videos.
• Site made $60K/year; sold on Flippa at 16 for $100K.
• Motivation to sell: met a 25-year-old who'd peaked in high school with a similar product and never moved on — didn't want the same fate.
• Reinvested proceeds into Cal AI; Henry (co-founder, also 17) and Blake Anderson (UMax, Riz GPT) joined as co-founders.

Why Cal AI

• MyFitnessPal is convoluted; existing trackers hadn't innovated since before AI APIs existed.
• Idea validated before a line of code: hundreds of health/fitness influencers already existed with audiences who would want exactly this.
• Confirmed the marketing channel first, then built the product.
• Tracks calories, carbs, protein, fat — deliberately nothing else.

Influencer marketing system

• CPM/RPM framework: only pay if projected spend per 1,000 views (CPM) is below projected revenue per 1,000 views (RPM). Start testing at a $5 RPM estimate; refine from there.
• Find influencers by creating a blank TikTok account, training the algorithm to the target niche, then logging creators into a Google Sheet (now an internal tool, soon viral.tech).
• Outreach via DM, not email — email routes to managers who charge more and negotiate less.
• Expect roughly 1 in 500 DMs to yield a usable creator.
• Bundling tactic: offer 4 videos upfront instead of 1; creators want cash, not video count, so bundling lowers per-video CPM and locks in recurring posting.
• Most common deal structure: monthly retainer for a set number of videos.
• Range: $100 to $100K per two months depending on scale and engagement.
• Engagement signals that matter: comments where users reply to each other, top comments with 100+ organic likes, no story-ring clustering (sign of comment-for-comment pods).
• Content brief: examples of past videos that performed well + 2–3 ideas. Creators do the rest.
• Only requirement: show the product in the first 15 seconds, never name it explicitly — let viewers ask in comments.
• Comment control: plant a pinned comment on every video; reply to every inquiry naming the app.
• All influencer videos reposted to Cal AI's own Instagram account; tens of millions of additional views per month at zero extra cost.
• Currently ~150 influencers posting on monthly retainers.

Attribution and tracking

• No link-in-bio clicks; users go directly to the App Store.
• Attribution method: overlay influencer post dates and view counts against App Store download/revenue graphs; isolate spikes to specific videos.
• This correlation logic is the core of viral.tech — the platform they're building publicly.

Pricing and monetisation

• Hard paywall: no free tier, no ads. $10/month or $30/year (most users choose annual).
• Premium-only model outperformed freemium with GrindClock — fewer downloads, more revenue.
• Three-day free trial only; paywall shown after onboarding.
• Profit margin: ~50%. Completely bootstrapped.
• Apple pays out two months after revenue is earned; all receipts to date reinvested into marketing.

Onboarding psychology

• Asks users: weight, target weight, height, goals, motivations.
• A/B tests revealed some questions have zero effect on the product — but asking them raises conversion.
• Mechanism: each question reinforces why the user needs the product; by the paywall they've made a mental commitment.
• App then sets a calorie and protein daily target; tracks cumulative intake vs. goal throughout the day.

AI and product accuracy

• First version: single ChatGPT prompt returning full macro breakdown — already ~80% accurate (matches FDA food-label margin of error).
• Current version: a pipeline of sequential, food-type-specific prompts reaching ~90% accuracy on average.
• Key user education gap: consumers expected X-ray vision; onboarding now includes a scanning guide to set realistic expectations.
• Tech stack: Swift (iOS native — smoother than Flutter/React Native), Firebase Cloud Functions (backend), Mixpanel (analytics).

Growth sequencing and scaling

• Sequence: build → App Store → DM 500+ influencers → get 2 posting → scale to 30+ → build content archive → layer in paid ads.
• Paid ads (Meta, TikTok, Apple Search Ads) started ~2 months in; spending $7K/day profitably.
• Influencer content doubles as ad creative — no separate production cost.
• Apple Search Ads insight: if bidding on a competitor's keyword is profitable, your product is meaningfully better.
• Reposting influencer videos to brand Instagram adds tens of millions of views/month; wait a few months before reposting the same clip.
• Next planned channels: TV ads, Google (currently no SEO investment).

App studio and viral.tech

• Cal AI's influencer network is the moat: new health/fitness apps can plug into 150+ creators immediately.
• Additional apps will bundle with Cal AI in influencer deals, increasing leverage and lowering CPM for both.
• Viral.tech (launching ~December 2024): go-to-market generator, influencer CRM, deal calculator, attribution dashboard.
• Rationale for building publicly: influencer marketing strategies are already being reverse-engineered; better to own the platform than defend the secret.

Mindset and long-term vision

• Cal AI is a stepping stone to financial freedom, after which the goal shifts to impact-driven projects at global scale (cites Elon Musk's Zip2 → PayPal → SpaceX/Tesla arc).
• Fears complacency more than competition — has seen peers with early momentum stagnate.
• College plan: gap year, city-hop (NYC, Miami, Bangkok, LA), test whether a social life is buildable outside campus; enrol only if it isn't.
• Key belief: inherent virality of an idea is the single most important filter when choosing what to build. If you can't immediately picture the influencer video, reconsider.
• Advice: start immediately regardless of age; tools like ChatGPT have eliminated most technical barriers to entry.