React Summit Asia 2027

Ask an AI agent to build a CRM, a project tracker, or an analytics dashboard, and something strange happens: the products start looking and behaving remarkably similar. This talk goes beyond screenshots to reverse-engineer why AI-built React applications converge on the same components, state patterns, architectures, and interaction models, and what that means for engineers building software with agents.

Our production Next.js micro-frontend app took 57 minutes to build. Using only Node's built-in CPU profiler, I found two bottlenecks, one in Module Federation and one inside webpack itself, and brought the build down to 26 minutes. You'll leave with a repeatable method for turning "the build is slow" into a hypothesis you can test.

“Your page loads are so slow.”Not an unfamiliar phrase for frontend engineers to hear. But how do we prioritise performance and tech debt when there are always features waiting to be built?A performance improvement becomes a business win when you can connect the engineering change to a measurable outcome.In this talk, we'll walk through how we proved that a 50% reduction in load times across our key pages led to 6% more trials started, from identifying the funnels with the highest potential return through implementation, instrumentation, data analysis, and the challenges we faced along the way.