AI drives nearly 30% of new code in software development

Source: CritpoTendencia Original Title: AI Drives Nearly 30% of New Software Development Code Original Link: A recent peer-reviewed study published in the journal Science confirms that generative artificial intelligence is rapidly transforming software development on a global scale.

In particular, the research led by the Complexity Science Hub shows that the proportion of code created with AI support has increased significantly in recent years, a change that is already beginning to reflect in both productivity and the economic dynamics of the tech sector.

Growing Adoption of AI in Programming

The analysis, which examined over 30 million contributions in Python made by approximately 160,000 developers on GitHub, shows that in the United States, the proportion of AI-assisted code rose from 5% in 2022 to 29% by the end of 2024.

Meanwhile, European countries like Germany and France are experiencing similar progress, with levels of 23% and 24% respectively, while India, Russia, and China also show increasing adoption, although access restrictions to advanced models influence these figures.

To reach these conclusions, researchers used an AI model capable of identifying whether code snippets were generated with tools like ChatGPT or GitHub Copilot.

Although adoption is expanding rapidly, regional differences remain evident and are largely driven by policies on technological access and the local development of advanced models, such as DeepSeek in China.

Experienced Programmers Gain More Benefits

In this context, the report provides a particularly revealing finding. While less experienced programmers use AI tools in 37% of their code, compared to 27% among more experienced developers, the latter achieve real productivity improvements, with an estimated increase of 3.6%.

Additionally, senior profiles tend to explore a broader range of libraries and tool combinations, employing AI not only to automate repetitive tasks but also as a catalyst for accelerated learning and technical innovation.

The Economic Value of Assisted Programming

From a macroeconomic perspective, the potential impact of artificial intelligence on the software sector is significant. In the United States alone, the industry allocates between $637 billion and $1.06 trillion annually to salaries related to programming tasks.

Under this scenario, if AI increases productivity by 3.6% and already influences 29% of new code, the annual economic benefit could be around $38 billion, according to the study’s estimates.

However, the authors warn that generative AI could widen the gaps between experienced and beginner programmers if access, training, and responsible use of these tools are not addressed.

Ultimately, the debate is not about whether AI should be adopted in programming, but about how to ensure its benefits are distributed equitably, avoiding further concentration of technological and labor advantages.