The expert predicted an increase in the use of AI to create code

By 2030, up to 70% of the software code in Russia will be generated by artificial intelligence (AI), predicts Mikhail Khlebunov, Director of Product at Servicepipe. He told Izvestia about this on April 17.

According to him, this does not mean that a person will completely stop participating in the software development process, but AI will radically change the workflow, turning "coders" into so-called "code architects."

"The number of vibe coders is growing rapidly in the Russian IT market. This is an approach to software development where, instead of writing code in the traditional way, the developer formulates a problem or desired functionality, and the AI generates the appropriate program code. One of the reasons is the reduction of IT teams that we have been seeing in the market lately, as the AI—created code is noticeably cheaper than the code written by a human," Khlebunov noted.

Currently, about 8-10% of lines of code in Russian systems are generated using AI, and by 2030 this share may increase to 70%. First of all, we are talking about solutions where high performance is not required, for example, the interface of websites and mobile applications, the expert noted.

In geometric progress: Russia should become a leader in the scale of AI application

At the AI Journey conference, Putin predicted a revolution in several areas and learned how GigaChat sped up the processing of direct line appeals

Moreover, in his opinion, such specialization as "code architect" will become popular in the future. We are talking about specialists who will design the internal system and the appearance of the IT product, but the writing of the code is completely delegated to GPT models.

However, at this stage in the development of generative models, AI-written code cannot yet compete with human-written code in terms of quality and security, Khlebunov believes.

"Vibe coding is now quite appropriate in the frontend, but this does not apply to the core of high-performance systems that process large streams of information in real time. The danger of vibe coding is that a developer who is used to assembling a product from pieces of code offered by GPT models will not be able to check and clean it of vulnerabilities that can lead to hacking of the system. In this case, only anti—malware bot protection systems will help protect the infrastructure from attacks, which prohibit them from scanning websites for "holes" in the code," the expert noted.

Earlier, on April 9, it was reported that more than 70% of IT specialists use artificial intelligence functions at least once a week. The main scenarios for interacting with such services are: working with code (73%), text generation (31%), information search (31%), data analysis (31%), creating presentations (24%) and image generation (24%). AI is less often used to analyze cybersecurity risks (19%), decode audio (15%), and prepare reviews of voluminous information (9%), such as meeting abstracts or analytical reports.