After the "One-Person Company Support Program" opened for submissions, we received a proposal that made us pause. Not a mini program.
A long md document — a cardiovascular and cerebrovascular disease smart collaborative management platform. It's meant to serve 24 township health centers, 11 in pilot. 9,000+ resident records running.
The AI + CloudBase projects we'd seen before were mostly mini programs: knitting, e-commerce, travel, games — all great, but all instantly imaginable "hobby projects."
This one is a genuine medical-industry informatization system — the kind of B2B project that used to require bidding, tier-3 security compliance, and a whole team grinding for over half a year.
Its author is Mu Ze. A 16-year medical informatization product manager who started his own "one-person company". The entire system was built by him alone with CloudBase + CodeBuddy.
We interviewed him. Afterward, one line kept turning in our heads —
So that's another way CodeBuddy + CloudBase can be used.
1. "What I can do to 90, I can usually only deliver to 60"
Mu Ze is 38. HIS, electronic medical records, family-doctor contracts, online registration, chronic-disease management, public health — he's done nearly every mainstream product in hospital informatization. In 2021 he committed to one direction: public health and health management.
The longer he did product management, the more suffocated he felt. Because many projects' problems weren't technical at all.
Contracts are won by sales and pre-sales; technical evaluation and requirement management are nearly blank. By the time the project manager receives it, the contract is signed and the requirements are vague. The delivery goal is "conforming to the contract", not "users actually using it". The system that finally ships, primary-care doctors never open.
He had better data models, more reasonable extensibility structures and more flexible underlying designs in his head. But they got discounted layer by layer in execution.
"What I can do to 90, I can usually only deliver to 60."
It's not just healthcare. Any product veteran who's spent over 5 years in a traditional software company likely resonates: the thing in your head and the thing you can finally deliver differ by 30 points.
Those 30 points aren't an ability problem — they're systemic loss.
In summer 2024, Mu Ze resigned. He held two things: a chronic-disease management plan he'd repeatedly refined in his head for years — a multi-level institution collaboration mechanism — and a few clients from his former company who trusted him. He opened a one-person company, planning to make up those 30 points.
2. A thing he'd thought about for years, always running in his head
How do you ask a village doctor who has to fill 8 kinds of forms at once to do chronic-disease management well?
The thing in Mu Ze's head was meant to solve exactly this. But he didn't plan to build a "chronic-disease management platform"; he wanted an underlying framework that could extend to any health-management scenario through dynamic configuration:
- Form structure, dynamically configured
- Validation rules, visually configured
- Metric definitions, follow-up flows, reminder rules, all engine-ized
- A new business doesn't need a from-scratch build — just maintain some data and it runs
In his own words: "I don't build one kind of disease; I build the soil that can grow all diseases."
He'd refined this plan in his head for years. But it always stayed in his head. Because in the old world, getting it out into the world required passing five gates: a full-stack team, a server cluster, an outsourcing contract, a year-plus schedule, and round after round of "user experience vs implementation cost" compromises.
He's a product manager. He has ideas, but not the "construction crew" to realize them.
3. From "make a demo" to "accidentally made a real product"
In November 2025, Mu Ze got a cardiovascular-platform contract. The original plan was to outsource to a 3-4 person team.
He organized the PRD, prototypes and business logic, and handed them to the outsourcer. It turned out the outsourcer had taken several projects at once, and his was deprioritized. After a month of talks, almost no real progress. He waited and fretted. The contract was his; the delivery date pressed on him.
Then he did something — he started making the demo himself. Not to write code, but to replace documents with a runnable prototype, reducing miscommunication with the outsourcer.
Around then, he saw a video on Bilibili. Someone was demonstrating writing code with AI, building a complete small project. It was late 2025, MCP was just spreading, and AI coding tools were becoming genuinely usable.
He half-believingly installed CodeBuddy and started trying. The first time it ran, he froze:
"I realized I no longer needed that outsourcing team."
Then — he dropped the outsourcing plan.
From December 2025 to April 2026, he alone built the system from zero to launch. On April 2, 2026, the system was formally delivered to the client. 11 clinics and 1 township health center in pilot, 9,000+ resident records. Counting from when he saw that first Bilibili video — a bit over four months.
4. Why CloudBase: like running water at home
When it came to tech selection, I expected a list of comparisons. A 16-year veteran comparing three clouds, five backend approaches and seven deployment flows is baseline equipment.
His answer was almost too simple: Serverless is the answer.
And CloudBase's serverless architecture matched his years of microservice thinking almost seamlessly — each cloud function is a service, and the business is naturally modular.
One more key point — he needed not just a web end, but a consumer-facing mini program for residents. If he chose another vendor's serverless, he'd have to build the WeChat mini program infrastructure himself, wire up auth, and handle data integration. Carrying all that alone is a headache.
"You don't need to dig a well. You just turn on the tap."
For a product manager like Mu Ze who "already had the house blueprints in his head", what he never lacked was the blueprint — what he lacked was a piece of land already wired with water, gas and network. CloudBase is that land.
5. His division of labor with AI: I design the architecture, AI writes the code
If CloudBase is the land and pipes, AI is Mu Ze's construction crew.
And this "one-person company" boss — holding the PRD, architecture diagram, schedule and acceptance criteria all at once. His words in the interview are the most valuable of the whole session:
"I design the architecture, AI writes the code. I'm responsible for thinking the business through clearly; AI is responsible for landing it exactly. It won't find the requirements annoying, won't suggest simplifying because the implementation cost is too high, and won't walk away six months later full of compromises."
This clarifies one thing — AI programming's biggest dividend isn't letting people who don't understand business write code; it's freeing people who understand business from being trapped by "can't write code."
So AI recovered those 30 points.
6. But there were several pitfalls
The journey wasn't all a success story. Mu Ze didn't hide it — he admitted the early period was almost "code hell."
The pits: AI's early code was structurally messy, changing one place broke three; after context loss AI started "free-wheeling", deviating from his architecture; the database schema changed repeatedly, making data migration painful.
His solutions — worth a look for anyone in B2B: document the architecture and feed it to AI; break tasks small, change one module at a time; nail down the key data structures first, then let AI act.
7. Four months later, 140 institutions in use
On April 2, 2026, the system formally launched. The first batch of 11 pilot institutions — township health centers, community health service centers, village clinics — started migrating their work from paper forms to the platform.
Before launch, Mu Ze brought the Beta to the client's site for a demo. Halfway through, the client decided on the spot, arranging people to integrate and demanding it go live as soon as possible.
"It fit their expectations very well."
A week after launch, he'd already shipped two minor versions. Each came from primary-care doctors' real feedback — someone said a flow was too convoluted, someone asked to add a field, someone raised a scenario he hadn't thought of.
The latest update: the pilot passed on May 8 and is now rolling out county-wide, with the county CDC + township health centers + village clinics totaling 140 institutions in use. Mu Ze has started mini program development, demoed the beta to the client on May 18, and expects to launch mini program 1.0 in early June.
By his own estimate as a veteran PM, at this scale a traditional development model would need at least half a year and a 3-4 person team.
What he actually used: one person, one laptop, four months. Plus CloudBase's resource fees and CodeBuddy's token fees. That's it.
8. "The time gap between big companies and small ones has been leveled"
Later in the interview, Mu Ze said something. If you read only one passage of this post, read this:
"The time gap between big companies and small ones has been leveled. Before, big companies had a whole team and small ones didn't. Now I have AI + CloudBase, and what I can produce is no longer as far behind what a big company's team produces."
Coming from an indie developer two or three years in, this would sound like youthful bravado. But from a veteran who's done 16 years of B2B and watched one "built but unused" project after another — it carries a completely different weight.
He continued: "The B2B threshold now shifts from 'can you assemble a team' to 'do you understand what users truly need.' Teams may disappear, but understanding users won't."
9. Epilogue — he's just the one walking ahead
Near the end, we asked: "16 years in medical informatization, up to this project, what's your deepest feeling?"
He thought for a moment and said: "So the things in my head really can be used by people."
It sounds light. But it needs three conditions to hold at once to be said —
- A product manager who understands the industry and business, with "it should look like this" in his head;
- A cloud platform that packages all infrastructure, freeing him from servers and ops;
- An AI that translates the architecture in his head into code exactly.
In the past, missing any one of these, the story couldn't be told. Now, all three are in place.
We wrote about Lao Mo before — 51, zero coding, a knitting mini program in 3 days. She proved: the barrier has collapsed, ordinary people can build things.
Today's subject Mu Ze is a 16-year medical informatization veteran. He proved: a veteran's accumulated experience can be reactivated by AI.
Who's next? A product manager who's done 10 years of supply chain, building an ERP for a vertical industry; a township veterinarian, building a livestock-health-record system for peers; a community grid worker, building a service-booking tool easier for the elderly.
As long as they've soaked deeply enough in their field and know "what users truly need."
