How to Create Your Own IT Project Using AI Without Coding Skills. Vibe Coding for Beginners. Introduction and Myths.

Introduction.

Introducing Aleksey – CEO of Riser

"I created Riser in three months; it’s my fastest-building project so far. The emotions are intense, and I experience daily interaction with AI alongside them. My thoughts on what has happened and what is happening.

The first version or MVP of the project was largely created with the help of artificial intelligence, and this series of articles will describe my experience gained from this and some other projects. Let me clarify right away, I am familiar with the terms “large language model” and “neural network,” and I know that technically they are not AI. However, for the purposes of these articles, I think it’s better to use the popular terminology.

For over 15 years, I have been managing IT projects in both large and small organizations, so my assessments of the prospects of various solutions are subjective but also backed by a certain level of expertise I have gained.

This series of articles aims to help people with great ideas but lacking the skills or inclination for programming to realize them and make the world a little better. If you consider your idea harmful to society, these articles are not for you.

A Bit About Vibe-Coding and Myths Surrounding It

Recently (the article is written in May 2025), the concept of vibe-coding or creating projects without coding knowledge is becoming increasingly popular. At the same time, I often see significant resistance from programmers, as well as from people who have tried to do something with AI but failed. I want to stand on the side of AI and challenge some prejudices based on my own experience:

Myth 1: AI produces terrible code

Current models from leading providers, such as OpenAI’s GPT-3 (ChatGPT) and Claude 4 (Anthropic), already generate code quality on par with the top 10% of programmers participating in competitions and olympiads. From my own experience, I’ve repeatedly seen that AI can optimize code and often suggest solutions I didn’t even think to ask for—pleasantly surprising!

Of course, if you try to build a project using something like the free version of ChatGPT or other simple models, you might waste a lot of time and get terrible code. I recommend not skimping on the reasonably accessible prices—$20 (or from $1 with Riser) for access to the best AI models.

Myth 2: AI will never replace an experienced programmer

It will, and quite convincingly. The question is about the tasks. For complex systems that are 10 years or older, where you need to know all their nuances and tuning subtleties, as well as coordinate with other developers and systems, AI currently cannot replace a human programmer.

However, for new projects or integrations, AI can program dozens of times more efficiently and a thousand times cheaper. For example, integrating a payment gateway with Stripe and Cloudpayments to enable card payments on a website took one day of work by one person and $3 worth of AI processing with Claude. Normally, that would take several days, involving reading documentation, debugging, browsing relevant sites, etc.

Having worked in product management at large financial companies, I was shocked in the beginning at how functionality that would normally take two weeks with two mid-level programmers can be done in a day using AI.

Conclusion: all soft skills—team communication, overall project management, coordinating business requirements, even casual coffee chats (I admit, I loved coffee, what can I say)—these will still be handled by programmers.

And their number probably won’t decrease; their skills will just evolve, and their efficiency will multiply.

Myth 3: AI often goes in circles but doesn’t produce a working solution

AI has one main feature: it works only with the data or context it’s given. I’ve experienced moments when a whole day was spent on what seemed like a simple task. Usually, it turned out that I provided insufficient data or that it was following outdated instructions. This isn’t a flaw of vibe-coding; it’s just a feature. But it’s not critically problematic. I will revisit how to avoid or minimize this in the next article, where I’ll describe specific working techniques.

Myth 4: Project creation is for professionals; with zero knowledge, there’s nothing to do

In vibe-coding, the difference between an experienced product manager and a beginner entrepreneur who wants to create a super idea that will conquer the world is mainly the number of AI queries. You’ll need to ask more questions and iterate more, but the achievement depends solely on persistence.

Myth 5: You need lots of money to work

Regarding the high cost of AI services, it’s a relative concept. Is it more expensive than free? Absolutely. Cheaper than hiring specialists and paying for their time? Hundreds of times cheaper.

It’s also important to understand that most AI creators, like OpenAI (ChatGPT), Anthropic, Deepseek, and others, currently operate at a loss. They are trying to capture the market and train their models to outperform competitors. Prices will rise in the future. So, create now.

The next set of thoughts will cover the first steps, common mistakes when working with AI, how to avoid them, and how to choose the right models."