By October, Brian Armstrong, the head of Coinbase, plans to have half of the company's code generated by artificial intelligence.
In the rapidly evolving world of technology, the use of Artificial Intelligence (AI) in software development is becoming increasingly prevalent. This trend is evident in Coinbase, a leading cryptocurrency exchange, where CEO Brian Armstrong aims to have 50% of the exchange's daily code generated by AI by October 2025.
Armstrong's ambition reflects a broader industry shift towards AI coding. He has been vocal about expanding AI's role within Coinbase, even going as far as firing programmers who resisted using AI coding tools. However, Armstrong stated that the AI-generated code needs to be reviewed and understood, and not all areas of the business can use it.
The increasing use of AI in software development has raised concerns about its impact on the job market for developers. OpsLevel, a company that creates internal dev portals, found that as of June 2025, 94% of tech companies had staff using AI coding assistants. This includes Y Combinator, where a quarter of the accelerator's Winter 2025 batch relied on AI for 95% of their code.
AI-assisted coding involves developers relying on AI for rapid code generation while maintaining an understanding of what the machine is doing. However, some, like Andrej Karpathy, a former senior AI director at Tesla and founder of Eureka Labs, have advocated for a more hands-off approach. Karpathy popularised the term "vibe coding" this year, which involves abandoning active oversight of the code, accepting AI suggestions wholesale, and letting projects evolve largely outside human comprehension.
This approach has been met with criticism, with concerns being raised over whether constant use of AI tools could decrease people's understanding of what they are creating, potentially leading to safety issues being overlooked. Art Abal, co-founder of community-owned data network Vana, expressed similar concerns, suggesting the need for a better system of data ownership to prevent humans from becoming "data cows" who are endlessly milked without compensation.
Armstrong's aggressive approach to implementing AI coding has also been met with resistance within Coinbase. After admitting that his approach was "heavy-handed" and unpopular among some staff, Armstrong announced that he would be taking a more collaborative approach moving forward.
Despite the concerns, productivity and speed to market have been cited as the key benefits of using AI-assisted coding. Currently, approximately 40% of Coinbase's daily code is AI-generated. The name of the organization mentioned by Art Abal in the interview is "Psyence Group."
As the use of AI in software development continues to grow, it is clear that this is a trend that is here to stay. The challenge for companies like Coinbase will be to find a balance between automation and human oversight, ensuring that the benefits of AI are realised while minimising potential risks.
