Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
ReleaseInsight · On RadarHabr AI→ original

A Clinic Replaced a 4-Million-Ruble Subscription with a Vibe-Coded App One-Third the Cost: What the Business Still Has

A private clinic had been paying more than 4 million rubles a year for a white-label app, a DocDoc analogue. It hired two developers and, in three months, built a replacement with vibe coding for one-third the price, with no subscription. The debate on Habr showed that cloning a product has become cheap, but code was never the only cost. Support, liability for data leaks, legal compliance and marketing remain. That is why product protection should be sought in service, law and distribution.

AI-processed from Habr AI; edited by Hamidun News
A Clinic Replaced a 4-Million-Ruble Subscription with a Vibe-Coded App One-Third the Cost: What the Business Still Has
Source: Habr AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceHabr AI ↗
Editor-in-Chief
Clinicover 4 million rubles a year for a subscription versus a replacement at one-third the cost
Torvaldscompilers brought a 1000x productivity gain, AI may bring 10x
2016 authorafter their small program was cracked, sales grew thanks to free advertising

A private clinic had been paying more than 4 million rubles a year for a white-label app, a DocDoc analogue. Updates kept coming, but customization was limited. So the clinic hired two developers, and in three months they built a replacement with vibe coding. It came out at one-third the cost. The clinic fully owns the app, can extend it without extra fees, and pays no annual subscription.

The story was shared in the comments to an article on Habr. The community split at once. Some believe the work of these two is worth more than the clinic is willing to pay. Others ask who will handle a personal data leak and fix a bug three years from now, when the developers have left. The argument remains unresolved. But it clearly shows how the economics of software are changing.

The moat around SaaS has shrunk

The core thesis is blunt: any software idea can be copied if it contains no know-how. A product takes six months to build and a week to copy. One Habr author frames this as an economic mechanism. Previously, a product was protected by the cost of replication: you needed a team, time and money. If a model takes over most of that work, the protection weakens.

The clinic illustrates this with numbers. The subscription made sense as long as hiring two strong developers for three months cost more. Now, for a narrow task with clear requirements, building in-house turned out to be cheaper. For SaaS founders, the conclusion is unpleasant. If your value is a set of standard screens and CRUD logic, a client can compare your price with three months of work by two people.

What a clone does not replace

The claim that cloning takes a week has weak spots, and the commenters named them themselves.

First, marketing. A product can be built quickly, but finding paying customers takes longer, and while you are still scaling, someone may already have cloned you. Second, brand and service. People pay for the original because they know it will not fall apart in a week and comes with a warranty and support. One author recalls 2016: his small program was cracked and posted on a forum, but sales did not fall. They grew thanks to free advertising. Users needed technical support, not just a file.

Third, law. For a replacement such as 1C, the value of new software is close to zero until you can guarantee that calculations and reports comply with current legislation. And fourth, knowing what to copy in the first place. One comment in the debate is accurate: if "any fool" knew which service was worth replicating, they would have done it a year ago by hand.

We have been through this before

A skeptical historian points out that the promise that "everyone will be able to program" has been around for a long time: COBOL in 1959, the Logo turtle in 1967, the Soviet DRAKON language for the Buran shuttle in 1986, Visual Basic in 1991, then low-code, and now vibe coding. Each time, the tool removed part of the routine, but deciding what to build and whether it was built correctly still fell to a human. Assembling an app became easy, but understanding what is inside it is harder.

In a debate about AI in the Linux kernel, Linus Torvalds said AI would raise productivity 10x, while compilers raised it 1000x. The world did not wipe out the value of programming after compilers appeared. This is no guarantee for the future, but the precedent argues against panic.

There is also a less pleasant version of the thesis, about the profession. A developer's good life partly rested on information asymmetry: a manager could not verify how long a task really takes, and "it's complicated" ended the conversation. AI agents force knowledge to move from people's heads into tickets and documents that managers can read too. One author predicts a split: a few expensive experts at the top, many cheap executors at the bottom, and a disappearing middle.

Practical takeaways

For teams and founders, this debate suggests several rules.

  • Do not treat code complexity as protection. Build your moat in data, integrations, service, licensing and legal compliance.
  • Count the cost of owning a clone honestly. A subscription covers not only code but also responsibility for support, security and updates.
  • If you are a client building a replacement, budget for maintenance. The question "who will fix it in three years" stays unanswered until the company has someone who has read this code.
  • If you are a developer, your value shifts toward framing problems, verifying results and taking responsibility for them.

An honest caveat: the clinic case is a single example retold in the comments, without verified figures on support costs. It illustrates a trend but does not prove that subscription products will disappear. Code has become cheaper, that is a fact. But the business around code never rested on code alone, and the value has moved to where an AI agent helps least: choosing the idea, earning customer trust and bearing legal responsibility.

Story

Key Questions Answered

What are the key details on The moat around SaaS has shrunk?

The core thesis is blunt: any software idea can be copied if it contains no know-how. A product takes six months to build and a week to copy. One Habr author frames this as an economic mechanism. Previously, a product was protected by the cost of replication: you needed a team, time and money. If a model takes over most of that work, the protection weakens. The clinic illustrates this with numbers. The subscription made sense as long as hiring two strong developers for three months cost more.

What a clone does not replace?

The claim that cloning takes a week has weak spots, and the commenters named them themselves. First, marketing. A product can be built quickly, but finding paying customers takes longer, and while you are still scaling, someone may already have cloned you. Second, brand and service. People pay for the original because they know it will not fall apart in a week and comes with a warranty and support. One author recalls 2016: his small program was cracked and posted on a forum, but sales did not fall.

What are the key details on We have been through this before?

A skeptical historian points out that the promise that "everyone will be able to program" has been around for a long time: COBOL in 1959, the Logo turtle in 1967, the Soviet DRAKON language for the Buran shuttle in 1986, Visual Basic in 1991, then low-code, and now vibe coding. Each time, the tool removed part of the routine, but deciding what to build and whether it was built correctly still fell to a human. Assembling an app became easy, but understanding what is inside it is harder.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…