Meta CEO Mark Zuckerberg recently shared his vision for the future of AI in software development, predicting that within the next 12 to 18 months, most of the code for the company’s Llama project will be written by AI. In a podcast interview with Dwarkesh Patel, Zuckerberg explained that AI is already at the level of an "average very good person" on a development team, but he anticipates that soon AI will surpass even the top human coders in its capabilities.
Currently, AI is proficient at auto-completing code, assisting developers by suggesting sections of code as they work. However, Zuckerberg's expectation is that AI will soon be able to handle more complex tasks, such as running tests, identifying bugs, and writing high-quality, independent code. This will significantly transform how software development is approached within Meta.
Zuckerberg went on to elaborate that AI is deeply integrated into Meta's toolchain, where it is being developed specifically to assist in the Llama research project. Meta focuses on creating a coding agent and an AI research agent tailored to its internal needs rather than building a generalized development tool for external use. He emphasized that this toolchain integration would become a crucial part of Meta's operations, ensuring that AI plays an increasingly central role in their development processes.
Earlier in the year, Zuckerberg expressed similar views, stating that AI could eventually replace mid-level software developers. This idea aligns with predictions from other tech leaders as well. Dario Amodei, CEO of Anthropic, echoed Zuckerberg's vision in a March 2025 interview, predicting that AI will generate 90% of the code within the next three to six months, and by the end of 2025, AI will be responsible for 100% of the code.
In line with these predictions, Google CEO Sundar Pichai recently revealed that 25% of its code is already being written by AI, while OpenAI CEO Sam Altman noted that AI generates 50% of the code for some companies. These growing AI capabilities in coding are becoming a significant trend across the tech industry, reshaping the future of software development.
