May 9, 2025, 10:00 AM ETInterview, Podcast

How AI Coding Agents Will Change Your Job

Adoption rates among Y Combinator founders for AI-driven coding have surged from approximately 0% (two batches ago) to 25% (last batch) and now 33–50% (current), with some teams writing the majority of their code using these agents.
Individual non-developers with limited professional coding experience can now generate substantial software artifacts, demonstrated by the creation of "RecipeNinja.ai" (35,000 lines of code, interactive voice agent, thousands of users) and a personal blog migration (15 years of posts, new hosting, new software) in roughly 90 minutes.
Productivity gains for users with basic technical knowledge are estimated at 10x compared to their performance at the peak of their professional software engineering careers a decade ago.
Tom Blomfield argues that the "combine harvester" analogy applies to software engineering: while food production increased 10x, the number of humans required per unit of food dropped roughly 1000x, a trend expected to be even more pronounced in software.

Blomfield predicts that traditional software engineering jobs will not exist in five to ten years, replaced by roles focused on "wrangling" AI coding machines and managing high-level agency rather than writing syntax.
The argument that AI will never be capable of handling professional codebases is dismissed as a losing proposition, given the rapid trajectory of model improvements and tool form factors.
Despite the potential for unlimited demand for software (the "electricity paradox"), the consensus is that AI will fulfill the vast majority of this demand, leading to a net reduction in human software engineering roles.
The future of software creation is expected to shift toward "ephemeral," on-demand custom code generated to solve specific user problems rather than static, long-lived applications built by large teams.
A key divergence in the future human-AI dynamic is the lack of AI "obsession" with product excellence; successful products will likely still require a single human leader to obsess over the user experience, a trait difficult to encode.

Industries such as law, medicine, and accounting are transitioning from viewing AI as a fringe experiment to a competitive necessity, with regulatory bodies and professional unions expected to create protectionist barriers (e.g., banning AI from prescribing drugs) despite the technology's superior capability.
Founders are already succeeding in these previously resistant sectors (e.g., Agora in legal), driven by the market reality that failing to adopt AI will become a competitive disadvantage within one to two years.
The cost of knowledge work is projected to decrease significantly, creating massive consumer surplus but also posing risks of mass displacement for white-collar workers compared to physical laborers like surgeons or electricians.
Societal turbulence is anticipated over a 10 to 20-year transition period as hundreds of millions of workers face displacement and the difficulty of retraining into new roles.

The current era is identified as the most advantageous point in history to found a software company, as individuals can achieve significant revenue (millions to hundreds of millions) with minimal capital, smaller teams (2–4 people replacing 40), and rapid development cycles.
Founders are advised to prioritize two core skills: staying updated on the latest AI tools to leverage a multi-year advantage, and developing the ability to identify and understand deep human problems, as technical execution will become commoditized.
Smaller, single-owner product teams are expected to produce higher quality design and user experiences by removing the organizational friction and unclear ownership often found in large, multi-team corporate environments.
Long-term optimism exists for a future of abundance (cured diseases, efficient systems), though the immediate transition phase involves significant risk of social and economic disruption.