Solo Dev + AI Mini-Program Hits Break-Even After 77 Days: Net Profit 2.06 Yuan

签小签 · Break-Even Retrospective & an Answer to “Can a Personal Mini-Program Actually Work?”

---

Conclusion first, no suspense:

77 days after launch, cumulative revenue is 131.06 yuan. Costs are 129 (server 99 + certification 30). (The domain is one I registered elsewhere; no extra cost, so I ignore it.)

Today I earned back the last 2.06 yuan. Broke even.

This means one satisfying thing: all costs are now wiped out. From this last cent onward, every yuan earned is pure profit. The server is no longer something I pay to keep alive; it's earning money back and feeding me.

In the previous post I said I'd recover the remaining 61 yuan before the next retrospective and officially declare break-even. A pile of commenters told me I was dreaming. That dream has come true now. The number is still too shabby to show off: 131 yuan won't even buy two people a decent meal. But behind those 131 yuan, the curve from loss to break-even to profit was scratched out cent by cent, by one person + AI.

---

The Ledger First

Three and a half months, four retrospectives, one curve that keeps climbing:

Stage	Days	Cumulative Revenue	Daily Avg
Half month	15 days	10.53	0.70
Full month	31 days	36.21	1.17
Month and a half	44 days	67.41	1.53
Break-even	77 days	131.06	1.70

Over the 33 days from 09/22 to 10/25 I earned 63.65 yuan, averaging 1.93 per day. In the last post I predicted “less than a month to recover the remaining 61 yuan” — it took 33 days, off by three. Honestly, that precision is a bit frightening: it means the rules behind this system can genuinely be mapped out.

The full-cycle ad funnel looks like this:

Pulls	Impressions	Impression Rate	Clicks	CTR	eCPM	Revenue
45,186	24,115	53.37%	281	1.17%	5.43	131.06

Same old line: not a cent spent on promotion, all organic traffic.

---

A Counterintuitive Thing: the Money Is Getting “Cheaper”

I pulled out the last 7 days of data separately:

The impression rate climbed from the early 48% to 57.85%. Ad placement tuning is paying off: the same volume of traffic yields more monetizable impressions. That's good.

But look at eCPM: it dropped, from 5.43 over the full cycle to 4.02 over the last 7 days. Click-through also fell from 1.43% to 1.13%.

In plain language: more people are viewing ads, but each impression is worth less.

I panicked at first, thinking I'd done something wrong. Then I worked it out: this isn't a product problem, it's the normal state of traffic. Early traffic was all precise users who searched the long-tail keyword “观音灵签” (Guanyin fortune sticks). Strong intent, high clicks, naturally better unit prices. Now the base is bigger, more general traffic has mixed in, and eCPM dilution is inevitable.

Raising the impression rate is something operations can control; eCPM is given by the market, and what you can't control you accept. As long as the total revenue line keeps climbing, I don't panic over unit-price fluctuations.

---

Revenue Trend: the Curve Finally Stops Looking Like an ECG

Back in August the curve basically hugged zero, occasionally hopping to 2 yuan and falling back; it was hard on the heart. In September it slowly held at 2 yuan. October still bounces up and down, but the center of gravity has clearly lifted.

The single-day peak is just over 4 yuan, with 1 yuan as the floor. It's no longer that “3 yuan today, 0 yuan tomorrow” rollercoaster; there's a faint gravitational line. In three months as a Traffic Master, the biggest mindset change lives in this chart: from staring at numbers daily in anxiety, to watching weeks and trends.

---

An Embarrassing Admission: I Actually Ran the DAU Group

In the last post I said I'd pulled together a group, gathering people willing to give a daily shake, and I'd share a bit of the earnings to buy coffee. The comments were a chorus of “that works?” and “isn't that just farming DAU?”

Here's the status of this “shameless” little loop:

• 20-odd people in the group, all friends who genuinely use it;
• 5 Luckin coffees drawn as of today, at a pace of at least one per week, promises kept;
• The group charges nothing and runs no tricks; cups of coffee hold together the understanding of “you help keep my DAU, I split you a coffee.”

Someone asked me whether this counts as self-deception. I don't think so. The 20-odd people in the group are real users who genuinely open the mini-program every day, genuinely shake for fortunes, genuinely watch ads. Machine traffic is fake; these people's DAU is real. I make the product, they spend time, I share the revenue: an honest little loop that's a bit plain.

The effect shows up directly in the data:

Cumulative users: 4,378. Daily visits: 52. Daily opens: 102. The key is that “102 opens / 52 visits” ratio: each person opens the app nearly 2 times a day. Someone who shakes and leaves wouldn't manage 2 times. That extra open is repeat visits fed back by the Luckin coffee and the group.

The Traffic Master business, plainly put, runs on “people coming every day.” The DAU leveraged by those 20-odd people is the confidence behind my sitting here writing the word “break-even” today.

---

OK, Back to the Real Question: Can a Personal Mini-Program Work?

Across three retrospectives, the most asked question has been: “Can one person really make money with a mini-program? Is it worth it?”

I'll put the most honest data on the table first, no motivational fluff:

What did I actually invest? I went through my git history:

• First commit 07/20, last commit 10/21, spanning 3 months;
• But the days I actually wrote code or changed things add up to only 31 days, with 77 commits total;
• The development phase was really just the three weeks from 07/20 to 08/10; the remaining two months were piecemeal maintenance.

What did I earn back? 131.06 yuan. Cost 129. Net profit 2 yuan 0 mao 6 fen.

Return on investment? If you treat “making money” as the sole goal, this project bleeds: 31 evenings of spare time for money that doesn't even cover a meal.

But measure with a different ruler, and the conclusion flips completely:

With 129 yuan and 31 evenings, I ran through something most programmers never independently run through in their entire careers:

One person + AI, building from zero a real product that's live, has real users, and can earn enough to support itself.

From requirements, UI, frontend, backend, deployment, domain, HTTPS, WeChat certification, Traffic Master integration, data retrospection, user operations... every pit on that entire chain I stepped into by hand and climbed out of. On the open market, a paid bootcamp quotes several thousand yuan for this experience, and it may not even teach you how to land it.

So my answer is: it works, but you have to pin your expectations to the right ruler.

• Want to earn 10,000 a month from it, feed a family as a side hustle: don't do it. You'll question your life in month three staring at two-digit income.
• Want to spend the price of one hotpot meal and a few dozen working days to buy proof that “I can independently launch a product and run the commercial loop closed”: absolutely worth it. It might be your highest-ROI self-investment this year.

The money broke even; the experience came free. That's the correct way to open a personal mini-program.

---

Oh Right, I Made a New Product with the Same Playbook

Once 签小签 proved out, I built a new mini-program using the same “person + AI” route:

厨菜记 (Chu Cai Ji)

A kitchen notebook that records “what you ate today, what to cook tomorrow.” The reason is mundane: I cook for myself, and every time I get stuck for half an hour on “what to eat today.”

Right now I'm validating one thing: whether the whole “launch — plug in Traffic Master — sustain DAU — break even” playbook from 签小签 can actually be replicated. If it can, that means what I figured out isn't luck but a reusable method. I'll sync progress in the next retrospective.

Search 厨菜记 in WeChat and give me your feedback.

---

In Closing

Three and a half months, 131 yuan, officially break-even.

From “recover the 61 yuan before the next post” to actually doing it, the next goal I've set for myself is simple:

From the next yuan onward, it's all profit. The server freeloads; the year-round ad money I use to feed the next product.

If you're also hesitating over whether to make something solo, my advice is one sentence:

Don't ask whether it can make money. Ask first whether you're willing to spend 31 evenings turning the three words “can it work” into a definite answer by hand.

Broke even, so an extra Luckin coffee draw in the group tonight. Said, and done.

---

Entry Points, Take What You Need

① Shake for luck 👇 Search 签小签 in WeChat, shake every day

② Want to join the group to raise DAU together and share Luckin 👇 Message me privately, state your intent

③ Want to see whether the playbook can be replicated 👇 Search 厨菜记 in WeChat, help me watch the data

④ You also want to develop, or want to learn how mini-programs are made 👇 I started another discussion group for products, AI development, and stepping on pitfalls. If you want to build one yourself, or you're curious how this “person + AI” actually lands, message me to join the group, and let's tinker together.