How AI Coding Agents Will Change Your Job

Tom compares coding agents to combine harvesters: they can raise output while reducing the number of people needed for the work. He describes building a blog platform in 90 minutes and a 35,000-line recipe app without writing code, drawing on his past experience as a software developer. He says YC founders are adopting these tools quickly, with the host citing a rise from about 0% to 25% of companies using them for most of their code across two batches.

Tom argues that coding agents will improve, and the host says dismissing them as toys overlooks how quickly tools can advance. They discuss Jevons paradox: cheaper software could increase demand, but Tom expects AI to meet much of that demand. They imagine on-demand software made for individual problems and predict that software engineering jobs will look very different within five to ten years.

The speakers compare coding agents with earlier abstractions, such as programming languages that moved engineers further from machine code. They debate whether humans retain an advantage in choosing problems and judging quality. One speaker points to a founder's deep focus on user experience, while the host says that agency and product ownership can make this a strong time to start a company with less capital.

Tom says YC companies are finding customers in fields such as law, finance, and medicine, despite earlier doubts about whether those industries would adopt software. He expects AI use to become a competitive requirement, though the host is less certain about how soon that will happen. They also discuss regulation and professional groups as possible barriers to adopting AI, even when it performs well.

The speakers describe a future with broader access to services, alongside a potentially painful transition if many workers are displaced quickly. Tom advises people to keep up with AI tools so they can recognize when the tools become useful in their field. He also recommends learning to identify and understand human problems, and expects smaller software teams to build better, more cohesive products.