Mang-Git sits down with Matthias Wagner, founder of Flux.ai, to explore how AI is finally making hardware design as accessible as software development. Matthias shares how Flux uses a combination of LLMs and proprietary spatial models to let anyone design PCB boards through a simple chat interface – and why LLMs alone fall flat when it comes to physical, geometric reasoning. The conversation digs deep into the build vs. buy dilemma in an era of rapidly shifting AI tooling, from when to roll your own stack to when to trust a third-party abstraction. Matthias also makes a case for using high-level AI agents like Devin to stop chasing the latest models and start compounding on integrations instead.

1:07 — Intro & Guest Background

4:30 — The Origin of Flux.ai

6:51 — Build vs. Buy: The AI Stack Reality

9:54 — Abstraction Layers & Developer Tooling

13:34 — How Flux.ai Uses AI to Design Hardware

15:53 — Where LLMs Fail: Spatial & Physical Reasoning

17:30 — Feedback Loops as the Real Moat

26:37 — Deterministic Tools vs. ML Models: How to Decide

31:27 — Lightning Round & Wrap-Up

This episode was brought to you by Anvil [https://useanvil.com].