English Is the New Programming Language (And We Are the New Bottleneck)

Generative AI and AI coding agents have quietly made English the most widely used programming language in the world, allowing people who would never call themselves programmers to generate complex outputs.

So why does it feel harder rather than easier?

Because only half of the work got cheaper. Producing things is now nearly free; deciding whether those things are any good is not. Generative AI scaled our ability to generate code far faster than it scaled anyone's ability to verify it, and the gap between the two is on us. We have become the bottleneck – not because we are slow, but because judgment is not automated.

This talk explores what this paradigm shift demands of the next generation of computer scientists. The focus has moved from knowing how to make something to specifying what you want and determining whether you got it. Drawing on current industry challenges and advanced applications of machine learning and automated reasoning at AWS, the talk will examine how to bridge this verification gap. It considers why relying on automated generation can feel uncomfortable, as verification is the capability we acquired by doing the work ourselves, and how rigorous programme analysis can help us scale human judgment in an AI-driven world.