Uploaded February 2026 | Updated September 2026, 1 day ago
Local-first ideas promise speed and resilience, but they get complicated fast once you add mobile constraints, AI systems, and real security risks. In this episode of *React Universe On Air*, Ola Linczewska is joined by Kraen Hansen to talk through what actually happens when those ideas meet production reality.
Kraen shares lessons from years of working on Realm and MongoDB, including local-first databases, and sync systems, as well as his more recent work at ElevenLabs. The conversation spans local-first architecture, React Native trade-offs, voice agents built on top of speech and language models, and hands-on security stories: from reverse-engineering a banking app to uncovering an XSS vulnerability on a government website.
The episode also touches on voice cloning, fraud prevention, and what it means to build AI systems responsibly, alongside personal stories about turning a teddy bear into a voice-driven game for kids, and agents playing 20 questions.
Chapters:
00:00 Intro and background
02:30 Why local-first still matters
05:30 Synchronous data and developer ergonomics
08:45 Local-first tooling and React Native
11:20 Local data and on-device AI
12:40 Voice agents and ElevenLabs
15:10 Voice-driven teddy bear experiment
16:30 Voice cloning, fraud, and safeguards
23:40 Reverse-engineering a banking app
32:20 XSS on a government website
36:30 Government software and open source
41:30 Parenting, boundaries, and closing thoughts
Check out episode resources on our website 📚
Catch more React Universe On Air episodes 🎧 clstk.com/4qIC2c8
Sign up for our newsletter ✉️ clstk.com/3OExvK4
Follow us on X 🐦 https://x.com/callstackio
Local-first ideas promise speed and resilience, but they get complicated fast once you add mobile constraints, AI systems, and real security risks. In this episode of *React Universe On Air*, Ola Linczewska is joined by Kraen Hansen to talk through what actually happens when those ideas meet production reality.
Kraen shares lessons from years of working on Realm and MongoDB, including local-first databases, and sync systems, as well as his more recent work at ElevenLabs. The conversation spans local-first architecture, React Native trade-offs, voice agents built on top of speech and language models, and hands-on security stories: from reverse-engineering a banking app to uncovering an XSS vulnerability on a government website.
The episode also touches on voice cloning, fraud prevention, and what it means to build AI systems responsibly, alongside personal stories about turning a teddy bear into a voice-driven game for kids, and agents playing 20 questions.
Chapters:
00:00 Intro and background
02:30 Why local-first still matters
05:30 Synchronous data and developer ergonomics
08:45 Local-first tooling and React Native
11:20 Local data and on-device AI
12:40 Voice agents and ElevenLabs
15:10 Voice-driven teddy bear experiment
16:30 Voice cloning, fraud, and safeguards
23:40 Reverse-engineering a banking app
32:20 XSS on a government website
36:30 Government software and open source
41:30 Parenting, boundaries, and closing thoughts
Check out episode resources on our website 📚
Catch more React Universe On Air episodes 🎧 clstk.com/4qIC2c8
Sign up for our newsletter ✉️ clstk.com/3OExvK4
Follow us on X 🐦 https://x.com/callstackio
