Last year at Current London, we filled a room and had an honest conversation about AI writing code. At the time, it felt like the early days. A year later, the agents arrived, and they're not just suggesting code anymore. They're writing and maintaining codebases, pipelines, and increasingly making architectural decisions that used to belong to humans.
Adoption is racing ahead of trust. Most streaming teams are shipping AI-generated infrastructure code faster than they can meaningfully review it, and almost nobody has a clear answer for what that means in two years, whether for system reliability, for engineering skills, or for the engineer's job itself.

This session is not a keynote. It's a structured conversation with the people in the room - the engineers and architects actually living this shift in their day jobs. We'll talk about what agents are genuinely good at in writing and maintaining code (and where they still fail spectacularly), how teams are rethinking review and accountability, what skills still matter, and what happens when agents start writing infrastructure for other agents to consume.

Come ready to participate and discuss what's happening today, and what's coming.