When AI writes the code, what do JavaScript developers write?

AI can now generate components, APIs, tests, database queries, and sometimes entire features. If writing the code is no longer the hardest part of building software, what does the JavaScript developer actually do?

Our role is moving up a level of abstraction. Instead of describing every implementation detail, we increasingly define intent, provide context, establish architectural boundaries, review plans, and verify what AI agents produce. But this isn't simply a transition from developer to prompt engineer. The engineering work hasn't disappeared it has moved.

In this talk, we'll follow that shift through a practical JavaScript development workflow. We'll look at how specifications can become sources of truth, how MCP and agent skills provide specialized context, and how types, schemas, tests, and linters become guardrails that agents can continuously validate against.

We'll also look at the other side of this transition: AI isn't only helping us build web applications, agents are becoming users of those applications too. Emerging browser APIs such as WebMCP allow JavaScript developers to expose structured capabilities that agents can use directly instead of forcing them to interpret interfaces designed exclusively for humans.

The result is a new responsibility for JavaScript developers: designing systems that can be understood and operated by both humans and agents.

Attendees will leave with a practical mental model for AI-assisted development that goes beyond prompting: define intent, provide context, establish constraints, let agents implement, and verify the result.

AI may write more of our JavaScript. The responsibility for what reaches production still belongs to us.