Edge AI models are nothing new, but for a long time their capabilities fell short of expectations. They could chat just fine, but often lost track of instructions or couldn't use tools properly. That perception still keeps a lot of developers from taking edge deployment seriously. Recent releases like LFM2.5 or Gemma 4 families change the picture significantly. They bring proper agentic capabilities, while being less resource-hungry than older models. And all of that on device - this means no server, no per-token bill, no user data leaving the phone.

After this talk you will be able to:
1. Judge whether your use-case suits models capabilities.
2. Build a real on-device agent loop - model, tool calls, and structured output working together reliably.
3. Understand various tradeoffs and concerns when integrating edge LLM to your product.
4. Know when edge isn't the answer, so you reach for it makes sense.
5. Add it to a React Native app with react-native-executorch.

The React Native ecosystem is changing fast, and AI is now influencing everything from code generation and testing to debugging, tooling, and app architecture. This panel brings together voices from across the community to ask what that shift really means for everyday RN developers: which AI and agent workflows are already useful, where the hype still outpaces the reality, and what skills will matter most in the next wave of mobile development. Expect a candid conversation about productivity, code quality, maintainability, and the practical changes teams will need to make as AI becomes part of the React Native workflow.