Coinbase CEO Brian Armstrong has disclosed that approximately 40% of the daily code produced at the cryptocurrency exchange is now generated through artificial intelligence (AI). In a recent post on the social media platform X, formerly known as Twitter, Armstrong expressed his ambition to boost this figure to over 50% by October of this year.
This statement highlights a growing trend in the tech industry, where major companies are increasingly harnessing AI capabilities for coding tasks. Microsoft CEO Satya Nadella noted earlier this year that as much as 30% of Microsoft’s code is AI-generated. During a discussion with Meta CEO Mark Zuckerberg at the LlamaCon AI developer event, Nadella remarked that a significant portion of the code in their repositories is now crafted by AI.
Similarly, Sundar Pichai, CEO of Google, stated that more than 25% of the company’s code is written by AI. In a third-quarter earnings call, Pichai emphasized Google’s advantageous position in the AI landscape, citing their unique full-stack approach to AI innovation.
In his post, Armstrong acknowledged that while AI-generated code significantly contributes to development, it must be carefully reviewed and understood. He noted that not all business areas can utilize AI in coding, but emphasized the importance of employing these technologies responsibly and efficiently.
Coinbase is actively expanding its use and familiarity with large language models, which are AI systems designed to process and generate human-like content. The company identified that teams adopting these models more rapidly are focusing on developing frontend user interface features, handling less-sensitive data backend tasks, and enhancing their unit testing suites.
Coinbase attributed part of its increased AI adoption to the implementation of an AI-powered code editor called Cursor, which has facilitated an uptick in the use of AI for coding tasks within the organization. This move aligns with a broader industry shift towards integrating AI technologies in development processes, reflecting a significant evolution in the way software is created.
