Most of use reach for a frontier model by default and pay for it on every call, in latency, in energy, in cash, and in everything that leaves their stack. For most of those calls, a small local model would do the job.

RL Nabors, former Meta/React core team member and AWS alum, covers the vocabulary you need to reason about model performance (capability evals, golden datasets, LLM-as-judge) and walks through real cases: a local agentic harness replacing a frontier call, an in-browser moderation classifier defended with production-trace evals, and a generative summarization feature where the rubric turns out to be harder than the model. You'll leave with a framework for deciding when to choose large and off-prem or small and local models, and how to measure your way to the answer instead of guessing.

You will learn:

In State of Brownfield in React Native, we’ll look at how engineering teams can introduce React Native and JS into mature, native-only applications. The session walks through practical strategies for phased adoption, cutting-edge tooling, and effective architectural choices. We’ll dig into the real challenges that come with hybrid integration and wrap up with advanced areas like brownfield state and data exchange.

The talk examines how teams can bring React Native and JavaScript into an existing native-only tech stack. We’ll look at practical approaches for gradual adoption—ensuring that iOS and Android apps can continue to receive features and fixes while part of the development workload shifts toward React Native. The session highlights state-of-the-art tooling, integration strategies, and architectural patterns that make brownfield adoption more seamless, while also addressing the real-world challenges that commonly arise in hybrid environments. We’ll conclude by exploring advanced topics such as preloading, passing initial props from native to JS, and efficient bi-directional communication between native code and JavaScript

Agentic engineering speeds up development, but fast-paced changes make edge cases—like onboarding flows and error handling—easy to miss.
Using Metro’s custom resolver, we can now launch directly into specific test environments, like targeted edge cases or simulated errors, without cluttering our production code with feature flags.

A walkthrough of a real session where two developers and their AI agent teams collaborated on a complex architectural change — going
from a rough spec to 8 PRs, 4,500 lines of code, and a clean task board in a single day.

The talk covers how we used git worktrees as a shared context surface between two developers' Claude Code agents, spawned specialist
subagents (including a Redis expert that caught two critical production bugs during design review), ran parallel implementation
across multiple branches, and coordinated via a dedicated agent that polled for commits and monitored Slack and CI.

The focus is on the collaboration model — not the technical feature itself — and what the human's role becomes when AI agents can
read, write, test, and coordinate autonomously.

TC39 is working on two proposals, import defer and export defer, whose goal is to allow developers to extract better performance from native ESM, both in the browser and on the server. Let's go together through how they work, and when and where you'll be able to use them!

How we transformed a React monolith into multiple apps inside one codebase.

You always start small: one product, one team, one stakeholder. But once you succeed and need to scale, your original frontend architecture might not be ready for what comes next. As our Sales Process Orchestration Platform expanded to support multiple products and sales channels, our original monolithic application became a bottleneck.
In this talk, I'll share how we transitioned to a domain‑driven, multi‑app architecture while deliberately keeping a single React codebase. You'll see how we applied Domain‑Driven Design concepts on the frontend, structured clear boundaries between sales channels, and used Nx to orchestrate builds, enforce architecture rules, and keep teams moving independently without losing shared foundations.