Donato’s local AI journey
Usman Pirzada from AMD interviews me about Strix Halo, Toolboxes and the local AI community.
Notes & chapters
I talk about how experimenting with Strix Halo led to Toolboxes, repeatable benchmarks and a community building specialised inference engines. We discuss Vulkan, ROCm, llama.cpp, Helogen, Gufo and ROCmFPX; models from Qwen, DeepSeek and GLM; and what comes next for local AI.
- Welcome and Donato’s Strix Halo journey
- Building and sharing Toolboxes
- Community engines and optimisations
- Local models: Qwen, DeepSeek and GLM
- What’s next for local AI?
This episode is sponsored by AMD.