Mojo, Modular, and the Future of AI Compute with Chris Lattner | Live from AGI House скачать в хорошем качестве

Mojo, Modular, and the Future of AI Compute with Chris Lattner | Live from AGI House

This week, we’re kicking off the first-ever AGI House episode with a special fireside between Modular CEO and LLVM/Swift creator Chris Lattner and AGI House CTO Henry Yin, recorded live at the inaugural Modular GPU Kernel Hackathon. They dive into the vision behind Mojo, Modular’s new programming model for AI developers, and the mission to democratize AI compute. Chris reflects on his decades of work unlocking performance, from building the foundations of modern compilers at Apple to now reimagining the AI software stack for the next generation of builders. They unpack why current AI tooling falls short, the architecture shifts needed to unleash developer creativity, and what it really takes to make hardware-accelerated AI programming accessible to all. Plus, Chris shares his journey from CPUs to GPUs, what inspires him today, and how Modular is thinking about the future of open AI ecosystems. Chapters: 0:00 Introduction 0:57 Chris Lattner’s Path 7:03 CUDA's Limitations 9:26 What is Mojo? 13:30 Mojo’s Compiler Stack: The Secret Sauce 16:24 Solving the Hardware Portability Problem 20:38 Can Modular Auto-Optimize for New Chips? 23:38 Making Mojo Usable for Developers 26:20 How to Start Using and Contributing to Mojo 28:05 Using AI to Build This New Language