OpenAI’s Chief Product Officer (CPO) Kevin Weil claimed that AI could surpass human coders in competitive coding benchmarks by the end of this year. His remark came during a podcast with Avalon CEO Varun Mayya and YouTuber Tanmay Bhat on their channel Overpowered.
Reflecting on the rapid growth of AI in recent years, Weil highlighted how AI coding models have improved significantly over time.
The Question
In the podcast, Varun asked Kevin for a realistic timeline on when functional code might be fully automated. Citing insights from Anthropic, Varun mentioned that functional code is expected to be 99% automated by 2027.
Anthropic’s Claude 3.5 Sonnet is among the top-performing AI models in recent coding benchmarks alongside OpenAI’s GPT-4o and o1 model. Anthropic has also introduced Claude Code, an agentic coding tool that allows developers to delegate coding tasks directly from their terminal, aiming to automate parts of the software development process.
Previously, Anthropic CEO Dario Amodei stated, “If I look at coding, programming, which is one area where AI is making the most progress. What we are finding is that we are 3 to 6 months from a world where AI is writing 90% of the code. And then in 12 months, we may be in a world where AI is writing essentially all of the code.”
The Response
Responding to Varun’s question, Kevin said, "At the rate we are going, I would be surprised if it is 2027. I think it is going to be sooner." He explained that writing good code requires a certain level of reasoning, and as reasoning models improve, coding capabilities will naturally advance.
Weil further mentioned that there are approximately 40 million coders worldwide. OpenAI’s o1 preview was initially ranked around the millionth position in competitive coding benchmarks. When OpenAI launched its first coding model, o1, it was ranked roughly as the thousandth-best engineer globally. OpenAI’s upcoming o3 model, set to be launched soon, is ranked 175th in the world, Kevin claimed.
Lastly, OpenAI’s Chief Product Officer stated, "As we are starting to train successor models, they are already better, and so I think this is the year that, at least by the competitive coding benchmark, AI becomes better."
In addition to this, Kevin also shared his insights on AI's rapid growth and its implications, deep research, GPT-4.5's human-like qualities, the future of jobs, and the expected timeline for GPT-5. He also revealed his billion-dollar startup idea and discussed how AI could revolutionise education and democratise software creation.
