FREE CHAPTER · THE MONEY REPORT
Shape 1: One-job phone apps
A free chapter from The 7 shapes of AI money, built from 196 sourced receipts.
Summary on one page
The receipts show seven useful shapes. A shape is a repeatable way that a product or a person turns an AI capability into a buyer payment, a business saving, a sale, or a prize. It is a pattern to test.
| Shape | Best example in the receipts | Numbers to remember | The practical lesson |
|---|---|---|---|
| 1. One-job phone apps | Stella by Sarah Pearl | $300,000 reported monthly recurring revenue; MVP in about two days; two to three months for the full build | Pick one painful job. Put one input and one useful output behind a paywall. |
| 2. Voice-to-text products | Letterly by Anton | $250,000 reported monthly revenue | Speech is messy input. Clean text is a clear, repeatable output. |
| 3. Get-me-customers B2B tools | Postiz by Nevo David | $239,848 Stripe-verified MRR | Buyers pay when the product is close to leads, sales, ads, search visibility, or published content. |
| 4. Faceless AI content | Boring History and the faceless channel network by Adavia Davis | $40,000-$60,000 reported monthly revenue; dashboard and payout records reviewed by the source | The product can be a repeatable content pipeline and a channel, not an app. |
| 5. Services are fastest cash | 1SecondCopy by Nick Saraev | $90,000 reported monthly turnover; fewer than three operators | Sell a result while people still need setup, judgment, and delivery. |
| 6. Speed is normal now | GeoSports by Frank Michael Smith | Built in eight hours over a weekend; $40,000 reported monthly revenue | Speed reduces the cost of testing. It does not remove the need for distribution. |
| 7. Small AI tools sell | NameSnag by Josh Pigford | About 12 hours to build; $15,000 reported acquisition price | A small tool can be an asset if it has a clear job, a buyer, and evidence of use. |
The database is strongest when it gives a processor check, public record, or dated maker document. A maker claim without a dashboard, contract, payout record, or independent confirmation is weaker evidence.
Shape 1: One-job phone apps earn the most attention
What the shape is
This is the simplest product pattern in the set: one narrow user problem, one phone-friendly input, one useful AI-assisted result, and a paywall or subscription. The input may be a food photo, a body goal, a chart, a sports video, or a daily habit. The app does not need to be a general assistant. It needs to be useful at the moment the user opens it.
The highest visible numbers in this shape come from self-improvement. Cal AI reads food photos and estimates calories and macros. AbMaxx creates an abs plan. Erly makes the alarm stop after push-ups. ChartDetector AI reads a trading chart. The common point is not the model. It is a clear promise that can be understood in one sentence.
Matching receipts
| Product and maker | Amount, approximately USD | Build time | Tools | Evidence |
|---|---|---|---|---|
| Stella - Sarah Pearl | $300,000 reported monthly recurring revenue | MVP about two days; full build two to three months | Claude Code, Cursor, Google AI Studio | maker-documented |
| Cal AI - Zach Yadegari | $115,000 in its second month | not stated | OpenAI GPT models, Swift, Firebase | maker-documented |
| CalBuddy - Tomer | $81,000 reported monthly revenue | two weeks to a working version | Cursor, Claude | maker-claim |
| AbMaxx - Lino Leighton | $60,000 reported monthly revenue | second version about one month | Lovable | maker-documented |
| ChartDetector AI - Timo and Patrick Kohler | $56,000 reported monthly revenue | not stated | Cursor AI, OpenAI, RevenueCat | maker-claim |
| Erly - Jake Glacer | $50,000 reported monthly revenue | 19 days from first commit to App Store | Claude Code, OpenAI | maker-documented |
| Wrestle AI - George Lampropoulos | $131,863 reported total revenue over six months | about one month | Rork, ChatGPT, OpenAI | maker-documented |
| 3AK Track - Christian Rac and Braylin Byrd | $17,460 reported revenue in the last 28 days | eight to ten hours a day before launch | Rork, ChatGPT | maker-documented |
| Cat on Chair - Ryan Yao | $18,000 reported sales in a 30-day period | about three months to first version | ChatGPT | maker-documented |
| STOPPR - David Attias | $12,000 reported monthly revenue | about one month | Cursor | maker-claim |
| Pep AI - Cedric Roberge | $60,000 reported MRR | not stated | Replit, Claude | maker-claim |
What the stronger examples have in common
The products have a short path from action to payoff. Take a photo. Record a voice note. Scan a chart. Watch a wrestling clip. Set an alarm. The product then returns a result that feels personal enough to justify repeat use.
The pricing is not visible in every receipt, so do not invent a subscription price from the revenue. The important pricing fact is the use of a paid app or recurring revenue model. Distribution includes the App Store, niche communities, paid TikTok advertising, and founder-led promotion. ChartDetector AI reports $56,000 monthly revenue while also reporting about $18,000 a month in TikTok ad spend. Revenue without acquisition cost is not profit.
Build time is often measured in weeks. CalBuddy reports two weeks to a working version. Erly reports 19 days to the App Store. STOPPR reports about one month. Stella separates a two-day MVP from a two-to-three-month full build. That separation is important: a testable product and a finished product are not the same thing.
The receipts state little about build cost. That is a warning. A fast build can still have recurring model costs, app-store fees, support, refunds, paid acquisition, and content costs. The correct first test is not “can I make this?” It is “will a defined group pay to use this next week?”
Failure risks
- A narrow job can still be a crowded category. Calorie tracking, trading guidance, and habit apps already have substitutes.
- A personal result is not necessarily a reliable result. Health, nutrition, and trading products need careful claims and user expectations.
- App-store revenue screenshots do not show profit, retention, refunds, or acquisition cost.
- The product may be easy to copy. The defensible part may be the audience, distribution, workflow, data, or habit loop.
- A one-job app can become a feature, not a business, if the result is useful only once.
If you want to try this shape: a 5-step starting plan
- Choose one recurring moment, such as “I have a meal photo,” “I need a workout plan,” or “I cannot switch off my alarm.”
- Write the input, output, and user promise in one sentence. Remove every second job.
- Build the smallest version that produces the output and measures whether the user comes back.
- Put a real payment decision in front of a narrow audience before adding a large feature set.
- Keep the test if users pay and return. Stop or change the job if they only praise the demo.
Three under-served product ideas suggested by the data
- A photo-to-action tool for a neglected routine. The receipts show food, charts, and body plans. A useful gap may be a specific work or household routine where a photo produces a checklist, not a score. The reasoning is the same low-friction input, but the buyer is less crowded than calories or trading.
- A sports-specific review app for a sport below the obvious categories. Wrestle AI and 3AK Track show that users may pay for analysis and coaching notes. A new product should choose one sport, one video input, and one decision the athlete makes after the result.
- A habit app that produces proof, not motivation. Erly turns the alarm into a physical action. A new version could produce a simple daily record that a user can review or share. The test must measure repeat use, not just downloads.
Keep reading
The other 6 shapes, 21 product ideas and the full dataset are in the database: $29 →