Loading page…
Photograph a stock chart, get an AI read on where it's going — $56K a month from two brothers running TikTok ads 20 hours a month.
13–14K downloads/mo · $260K total revenue in 13 months · ~25% profit margin (April: $43.7K revenue, ~$20K ad spend, ~$11.5K profit)
"When you spend $1.50 per download and you earn an average of $3 a user, then you have a 2x ROAS. When you reach that level, then you have built a money printing machine and the main goal is to spend as much as possible in ads."
Fewer bars = easier, cheaper, or faster for an AI-assisted solo builder. Editorial judgments based on the case details.
Timo is 23 and was working a nine-to-five as a software engineer at a large German car brand. He and his brother had tried several businesses before; mobile apps were the first thing that worked. Thirteen months after launch, the app has done more than 90,000 downloads and over $260,000 in total revenue, currently running at $56,000 a month. He quit his job after six months. He now works 20 hours a month on it — per month, not per week — while travelling. The origin is a good lesson in riding an API release. When ChatGPT first shipped image upload, Timo built a Telegram bot that could analyse stock charts from a screenshot. That was the whole idea. Then, as he puts it, "we got the first initial idea to turn this Telegram bot into a mobile app." Pat Walls, the interviewer, notes the pattern immediately: a new capability lands in a big model's API and creates a short window where wrapping it in a consumer product is a real business. Our read: two things make this case unusually credible. Timo shows the revenue dashboard on camera, and he volunteers his cost structure unprompted — including the fact that his margin is 25%, not the 90% that app-revenue screenshots imply. That honesty is rarer than the revenue.
There is no organic phase in this business. Chart Detector AI began as a Telegram bot Timo built the week ChatGPT shipped image upload, and every user since has been bought on a single channel.
The quick-reference card — acquisition channels, replication playbook, risks & traps, and our verdict — is members-only.
Join the Unicorn Club — $5/moCancel anytime, in one clickPaste into Claude Code / Codex and get a working version of this product end-to-end.
Two production-grade prompts per idea — a full build spec and an SEO growth plan — plus 10 tailored build specs a week. Members only.
Join the Unicorn Club — $5/moCancel anytime, in one clickEvery claim traces back to a listed source — open any of them and check the number yourself. The exact words are quoted below, with a timestamp that jumps to the second each figure is said.
“We are currently at $56,000 per month, and we are getting around 13 to 14k in downloads every month.”
1:51 · hear it in the video“I recently did a tweet about this and I made in April $43,700 and we have a total profit margin of around 25% after all of our costs.”
5:57 · hear it in the videoData credibility: Starter Story founder interview with Timo, co-founder. He shows his revenue dashboard on camera and states the figures directly: $56K/month currently, 13–14K downloads/month, 90K+ downloads and $260K+ revenue over 13 months, and a full April breakdown ($43,700 revenue, ~$20,000 TikTok ad spend, ~15% Apple fees, ~$11,500 profit, ~25% margin). All numbers are founder-reported and unaudited; the dashboard is shown but not independently verified. Ad performance figures (the ~5M-view creative generating $15,000+) are also self-reported. Nothing in the video addresses the regulatory position of selling AI financial predictions to consumers — that analysis is ours.
The first dollar arrives within hours of the first ad impression, by design.
The flywheel here is financial rather than social. Once a campaign clears 2x ROAS the loop is simply to feed it: monitor weekly, raise budget while returns hold, and respect the ceiling — "don't increase your budget more than 20% every 3 days.
Two kinds of silence show up here.
Timo's own description: "It's AI chart analysis app. So you can just take a picture of a stock chart or a crypto chart and AI will analyze it and give you some directions how the stock chart will go." That is genuinely the entire product. A camera, a multimodal model call, and a formatted result. There is no portfolio tracking, no brokerage integration, no social feed. The app does one legible thing that produces an immediate, screenshot-able output — which is precisely why it works as a paid-ads product, because the value proposition fits in a three-second video. The pricing is the part worth stealing. Two plans: $12.99 per week, and $59.99 for six months. And a hard paywall — "we don't offer any free trials or something like that." Walls flags the six-month plan as something he has never seen before on the show and calls it genius. Our read on why it works: weekly pricing maximises revenue from impulsive, event-driven users (someone checking a chart before a trade), while a six-month term captures the committed ones without the psychological hurdle of an annual commitment in a category where nobody trusts their own attention span for a full year. We should also say plainly what this product is: an AI making directional calls on financial charts. That is a category with real regulatory and ethical exposure, covered in section 07.
Subscriptions, with a hard paywall and no free trial. Timo is explicit about why that combination is what makes paid acquisition viable: "That works the best in combination with ads because you get immediately your money back." A free trial delays revenue past the point where you can judge an ad campaign; a hard paywall closes the loop within hours. The numbers he discloses are the most useful part of the whole interview. Currently $56,000 a month, 13–14K downloads a month, $260,000 total in 13 months. And then the honest version — for April: "I made $43,700 and we have a total profit margin of around 25% after all of our cost. So the majority of the cost is the ad spend around $20,000 in ad spend and the Apple fees around 15% and made a total profit of around $11,500." So roughly: 46% of revenue to ads, 15% to Apple, the rest to tooling and taxes, leaving about a quarter as profit. Walls' framing is fair: "I think actually these numbers are pretty standard especially for people that are running paid ads." He also does the arithmetic that makes the headline title work — 20 hours a month against $11.5K profit is roughly $500 an hour. Our read: internalise the 25%, not the $56K. This is a media-buying business with an app attached.
One channel: TikTok ads. Timo gives a step-by-step playbook that is the most operationally specific thing in the interview. Get the app ready first — onboarding that sells the outcome rather than the features, heavy social proof, a hard paywall with no trial. He recommends studying competitors' onboarding flows with a tool like Screens Design. Set up tracking properly. TikTok needs to know who actually paid, which requires an MMP; he uses AppsFlyer and explicitly recommends hiring an Upwork freelancer to integrate it, "because it can be a really complex topic and there's a lot of stuff that you can misconfigure that in the end can cost you a lot of money." Then the campaign mechanics: at least six videos to start, because the algorithm needs material to test. Smart+ campaigns so TikTok handles audience and budget allocation. Optimise for subscription events, not installs — "a lot of people are trying to optimize on app installs. I would recommend to optimize on subscription events." Budget at least $50/day. Broad targeting, selecting only the country. And a counterintuitive tip: "A lot of people first want to advertise in the US. But what we saw when we tested a lot of ads in other countries is that you have a lot better return on investment in other countries." Then leave it alone for the roughly seven-day learning phase. Scale weekly on ROAS, never raising budget more than 20% every three days or the campaign re-enters learning. And keep a constant pipeline of new videos to outrun creative fatigue. His best single ad — just a screen recording of someone using the app — generated over $15,000 in revenue on nearly 5 million views.
Timo lists it directly. React Native + Expo for the app. Supabase for the entire backend. OpenAI for the model behind the chart analysis. RevenueCat for paywalls and subscription handling. AppsFlyer for TikTok attribution — which he calls "a pretty essential part." CapCut for editing ad creative. Cursor AI for coding. Currently around $18,000 a month going into TikTok ads. The cost structure, from the April numbers: ~$20K ad spend on $43.7K revenue, Apple taking ~15%, leaving ~$11.5K profit at a 25% margin. Note what is *not* a major line item: infrastructure and model costs never come up as a concern, which tells you the OpenAI bill is small relative to acquisition. In this business model, compute is rounding error and media buying is the P&L. Our read on the leverage point: at 25% margin, a five-point improvement in paywall conversion is worth more than any engineering work you could do. That is why Timo spends his 20 hours a month on creative and campaign monitoring rather than features. It is also why the tech stack barely matters — every choice he made is the boring, default option, deliberately.
The copyable core has four parts, and none of them are about stock charts. First, watch for new capabilities in the big model APIs and wrap them fast. This app exists because image upload arrived in ChatGPT and someone built the dumbest possible thing on top of it within days. That window recurs constantly. Second, build a product whose entire value fits in one photograph. "Take a picture of X, get Y" is the shape that makes paid social work, because the ad creative is just a screen recording. If your product needs explaining, the ad economics collapse. Third, the pricing architecture: hard paywall, no free trial, a weekly tier plus an unusual mid-length term. The six-month plan is the detail most worth copying. Fourth, the ads discipline: optimise for subscription events, six creatives minimum, $50/day, broad targeting on one country, don't touch it during learning, scale at most 20% every three days. What we would change: pick a subject with less regulatory and ethical exposure than telling people which way a stock will move. The identical machine — photograph a thing, AI returns a confident read, hard paywall, TikTok ads — works for plant disease, skin conditions on pets, car dashboard warning lights, mushroom identification, mould, wine labels, or sheet music. Choose one where being wrong costs the user nothing they can sue you over.
The subject matter is the biggest risk and the interview never addresses it. An app that gives "directions how the stock chart will go" is producing financial predictions for paying consumers. Depending on jurisdiction, that can touch investment-advice regulation; it certainly invites app-store scrutiny and user complaints when the calls are wrong, which they frequently will be. Anyone copying this in the financial space needs disclaimers, careful output framing, and legal advice — none of which appear in the transcript. This is the single strongest argument for applying the machine to a different subject. The structural risk is platform dependency squared. The business depends on TikTok's ad platform for demand and Apple's store for delivery. A change to either — ad policy, attribution signal, algorithm behaviour, store rules on financial apps — hits revenue directly and immediately. Creative fatigue is the operating risk Timo names himself: "when you spend too much money on one video that it will die because too many people saw your ad. So to prevent this, you always need to have a constant pipeline of new videos." The 20-hours-a-month lifestyle depends on that pipeline continuing to produce winners. And the margin is thin. At 25%, a rise in CPMs or a dip in paywall conversion can flip the whole thing unprofitable fast. This looks passive and isn't — it is a business that survives on continuous, competent media buying.
The four founder playbooks — first customers, first dollar, the flywheel, the silent stretch — are in the Unicorn Club, on this idea and on all 1,435 others. The story, the numbers, the sources and every later section stay open to everyone.
Join the Unicorn Club — $5/moCancel anytime, in one click