Every income figure shows its evidence grade and basis · Individual results are not typical
Project guide

Products shipped fast with AI coding

All 4 self-reported cases were built with AI coding, with income from 100,000 yuan in about six months to 232,000 dollars MRR; the ad-funded recipe mini program earned only a few yuan

Compiled from the 5 cases in this project: 4 grade B (founders' own accounts on Juejin, Indie Hackers and MRR Story, or a Business Insider report) and 1 grade C lead (Chucaiji, whose income figures contradict each other in the retelling and are not shown). On 2026-10-23 the project was renamed from 'AI-coded mini program (ad revenue)': none of the four qualifying cases is a mini program or lives on ads.

Compiled 2026-10-23

Key facts

4Cases graded A or B
1Grade C leads

Income ranges use cases graded A or B (grade B is the founder's own account, with no platform data to check it), in USD; grade C leads are listed but not counted

See the steps

How the cases did it, step by step

Each step names the cases it comes from. Public sources give no reliable time data, so there are no weeks or hours.

  1. Step 1 Pick a small problem everyone hasA messy desktop (Ferry), a Gmail inbox flooded by mailing lists (Trimbox), class-action settlements nobody claims (Payout), finding recipes and tracking the pantry while cooking (Chucaiji).Cases: Ferry (渡) · Trimbox · Payout · Chucaiji
  2. Step 2 Let AI write most of the code and ship in weeksNinjaPear's founder, at home with a broken foot, built and launched it in two weeks with Claude Code; Chucaiji went from zero to launch in 19 days; AI wrote about 70% of Ferry's code.Cases: NinjaPear · Chucaiji · Ferry (渡)
  3. Step 3 Bring in the first users with a free tier or your own audienceFerry lets users organise 50 files free, and the author's series of write-ups doubles as its acquisition channel; Payout's founders brought their own audience, and 80% of growth is organic.Cases: Ferry (渡) · Payout
  4. Step 4 Charge by usage or subscription, and know the cost per useEach Ferry run costs about 0.07 yuan in model fees, and the first order was 9.9 yuan; NinjaPear bills by credits with no minimum monthly fee; Trimbox and Payout charge subscriptions.Cases: Ferry (渡) · NinjaPear · Trimbox · Payout
  5. Step 5 At a certain size, consider sellingTrimbox began considering a sale at 232,000 dollars MRR and sold at the end of 2025 for more than 4 million dollars, using AI to model different valuation scenarios beforehand.Cases: Trimbox

At a glance

Ranges are the 25th–75th percentile of cases graded A or B, computed by script from the case data; both ends of a ranged figure are counted

Cases graded A or B
4
Monthly income (annual figures divided by 12)
$49,301 – $157,801

Skills involved

How many of the 4 cases graded A or B list each skill

AI-assisted coding3 of 4 cases

AI wrote about 70% of Ferry's code and 80% of its copy; NinjaPear was built and launched in two weeks with Claude Code; Trimbox used AI to turn React code into React Native for its iOS and Android apps.

Mobile app development4 of 4 cases

Payout is a mobile app that is not itself AI; AI went into building and marketing it, and it cost about 10 dollars to start.

Pricing1 of 4 cases

Ferry organises 50 files free, then charges to unlock; the first order was 9.9 yuan and each run costs about 0.07 yuan in model fees.

Short-video content & ads1 of 4 cases

Payout's founders brought their own audience and say 80% of growth is organic; they also use AI agents to comment under relevant social posts.

Mini-app & web building0 of 4 cases

Chucaiji, a WeChat mini program, went from zero to launch in 19 days with one person and AI; ad income after enabling WeChat ads was minimal.

Key points

How big the income is

The four qualifying cases are self-reported: Trimbox was at 232,000 dollars MRR when it considered selling and sold for more than 4 million dollars at the end of 2025; Payout reached 1,003,227 dollars in annual recurring revenue in nine months; NinjaPear passed 15,000 dollars of monthly gross revenue six months after launch; Ferry made more than 100,000 yuan in about six months. Chucaiji earned only a few yuan from ads, and the retelling's figures contradict each other, so it is kept as a lead.

  • A sale price is not revenue
  • Ferry's posts carry no product link, so its income is a floor
  • Turning on ads is not income; it takes heavy traffic

Where AI helps

More than code: AI wrote 80% of Ferry's copy; Trimbox used AI for design and a support bot, and to model valuations before its sale; Payout's product is not AI, but AI agents comment on social posts to find users.

  • Code and copy
  • Design and support
  • Acquisition and pre-sale modelling

Dependencies and risks

Trimbox depends on Gmail permissions and Google's review; NinjaPear is still in the data supply chain, and its founder's previous company, Proxycurl, was sold after LinkedIn sued it, so the legal risk remains; Payout's AI-agent comments risk platform bans; Ferry's feature is easy for an operating system to build in.

  • Platform permissions and review
  • Legal risk in where data comes from
  • Automated comments can get accounts banned

Cases graded A or B

See them all in the case library →

Other cases in this project

Grade C leads: income not reliably confirmed; useful for how things were done, but not counted above.

Tools

Tools and platforms the cases mention. Every link goes to the official site, and we earn nothing from them.

ToolUsed forLink
Claude CodeNinjaPear's whole platform was built with it in two weeksOfficial site ↗
React NativeTrimbox used AI to move its React code to it for the iOS and Android appsOfficial site ↗
Chrome Web StoreTrimbox began as a Chrome extensionOfficial site ↗
WeChat Official Accounts Platform (mini programs and ads)Where Chucaiji published its mini program and turned on adsOfficial site ↗

Platform rules and compliance

  • Apps that read users' mail or files must pass the platform's permission review (Google's review of Gmail scopes, for example) and publish a privacy policy.
  • If you trade in company or personal data, make sure the source is lawful; NinjaPear's founder's previous company was sued over LinkedIn data.
  • Using AI agents to post comments for acquisition can break platform rules and get accounts banned.
  • A service that touches legal matters should say what it is: Payout stresses that it is a discovery platform, not a law firm. Mini-program ads also require meeting WeChat's eligibility conditions.

Compare

FAQ

How much can a product built with AI coding make?

Our four qualifying cases are all self-reported: from Ferry's 100,000 yuan in about six months to Trimbox's 232,000 dollars MRR (later sold for more than 4 million dollars).

How fast can it ship?

NinjaPear launched in two weeks with Claude Code; Chucaiji took 19 days.

Can a mini program make money from ads?

Chucaiji earned only a few yuan after turning on WeChat ads; ad income needs heavy traffic.

Does the product itself have to be AI?

No. Payout's product is not AI; AI went into building and marketing it. Trimbox is an ordinary inbox cleaner too.

What risks should I watch?

Platform permissions and review (Trimbox depends on Gmail), legal risk in data sources (NinjaPear), and bans for automated comments (Payout).

Want more examples like these?

See them all in the case library →