AI, vibe coding, and the new developer market: what's happening in the industry

Boris Dipner, city — Karaganda, iOS development team lead at BCC Business, LinkedIn

About me 

I am a Senior iOS developer with over seven years of experience in iOS and eight years in IT. Currently, I work as a team lead.

I started with competitive programming in school, coding in C and C++. I engage in mentoring and participate in acceleration programs, and I am also one of the organizers of Google DevFest Kazakhstan and the iOS community — we have hosted over 20 meetups.

About 80% of my experience is related to FinTech; I have also worked in the marketplace sector, including Wildberries. In addition to development, I actively develop the IT community through events and mentoring.

What tasks can be optimized using neural networks

In work, this is primarily code review: I run an initial pass using AI, remove obvious errors, and then look deeper myself. This is especially useful when the same classes are used in different parts of the project — the neural network helps quickly highlight potential dependencies and affected modules. Documentation also goes there: it used to take a lot of time because you had to transition from code structure to a clear description of the domain. Now AI speeds up this process significantly — it can build diagrams and rewrite text so that anyone can understand it.

The same goes for tests: boring cases are now easily generated automatically, allowing you to spend more time on edge cases and complex business logic.

But the most interesting cases are pet projects. Recently, I had an experience where I did almost nothing manually for a couple of days: from idea to a ready application in the store. First, through Google Stitch AI and Claude, we designed the UX, user flows, and data structure, and then I generated the design.

Next, I connected tools via MCP and export from Figma. I use different IDEs — Windsurf is more convenient somewhere, Claude somewhere else, Cursor somewhere else. Through integrations, design and context become immediately available in the code. We also quickly discussed the architecture with Claude — it took about 20 minutes.

The result: design is ready, implementation is ready, tests are added — and the application can be published.

About the development and hiring market

The market has changed significantly. First of all, there are many influencers and developers who create their own products.

I, for example, had a simple idea: my eyes got tired, I forgot to put in drops — I only remembered when they started hurting. Plus exercises. I made an app like a fitness tracker, but for eyes: reminders for drops and fatigue. That was the whole idea.

And there are more and more such cases now — people are mass-producing applications and uploading them to the store. Because of this, even reviews have become longer: previously, a new application could be checked in a day, now it can take about a week. The market is simply overloaded.

The job market is also ambiguous. There are startups that raise investments and literally spend everything on AI tokens to experiment and see what can be automated.

And in hiring, the entry barrier for beginners has noticeably increased. Previously, you could not know something at the start — the main thing was that the person was reasonable, and the rest was quickly gained through experience. Now, due to widespread code generation, errors can be masked, and this becomes a risk.

Interviews have also changed: sometimes it is visible when a candidate uses AI in real time — answers sound monotonous, there are pauses, the feeling that he is just reading text. This significantly changes the dynamics of the interview.

I myself tried both sides — both interviewer and candidate with such tools in a test format. And I noticed that with a high technical base, this rather hinders than helps: it is easier to prepare honestly and pass the interview without crutches.

But there was also a good experience: one interview with a foreign startup where they gave 20 minutes and allowed the use of any AI tools. It was important not just the solution, but how you formulate prompts, how you work with tools, and make architectural decisions. We then analyzed the code and approach together — and this, in my opinion, is close to what future interviews will look like in the industry.

Recommendations 

People now learn more to pass interviews than to actually work.

I see this both in beginners and from my own teaching experience — I lead a group on iOS development, we even have a topic on vibe coding. And the market is already adapting: more live coding with verbal explanation, more practical tasks, less "tell theory from a book," more questions on understanding, not on memorization.

The main mistake of beginners now is using AI as a replacement for understanding. When a person simply receives an answer, copies it, and moves on. Yes, he can quickly "assemble" an application, but in fact, he only develops the skill of prompting. This is also important, but the foundation does not disappear anywhere: algorithms, data structures, understanding of the platform are still needed.

AI can generate code, but it does not explain why it is like that. And if you do not understand the basics, you simply will not be able to evaluate whether it did anything correctly at all.

Therefore, it is important to use AI as an enhancement, not as a crutch — when you already understand what is happening and can critically look at the result.

And separately — the skill of formulating tasks. A good prompt is indeed becoming an engineering skill now. Models do not "think" in the human sense, they predict the next token. Therefore, the key is to be able to clearly understand the problem and formulate the desired result.

Networking is one of the most important things for a developer. It works both ways: you can invite someone to a pet project, or you can yourself get into an interesting team through acquaintances. And when you have already worked with a person, it is easier for you to invite him to an interview or recommend him to a company later. Therefore, meetups and face-to-face meetings are a must. Even if there is no time for lectures or training, you can always get out for a couple of talks, meet people, and simply expand your network.

Where is it best to start a career  

I would recommend starting a career with a corporation — as a base for gaining experience.

In a startup or small projects, you can quickly hit a ceiling, especially regarding code quality and understanding of how it should look in production. And this is exacerbated if you actively use AI: it helps write code quickly, but does not always form the correct understanding of architecture and standards.

In corporations, processes are usually more structured. Not always perfect, but on average, there is more attention to code quality, review, testing, and engineering culture. Plus — access to people with more experience, from whom you can learn a lot.

Currently, the market in general is difficult: it is not so important where exactly to start — a startup or a corporation. If you manage to get a junior position, that is already good.

But the key — do not stop there. It is important to find a mentor, create your own development plan, delve deeper into technologies, and try to implement them independently, not only through AI. Then real understanding appears, of what you are working with and why solutions are arranged exactly like that.