I'm playing telephone pictionary with LLMs, VLMs, SDs, and Kokoro on my Strix Halo

An AMD engineer demonstrated a viral "telephone pictionary" chain using four AI models on the upcoming Strix Halo processor. The experiment sequentially passed prompts between large language models (LLMs), vision-language models (VLMs), image generators (SDs), and the Kokoro model. The demonstration showcases the chip's ability to handle multiple AI workloads simultaneously, hinting at future consumer devices running complex multimodal AI pipelines locally without cloud dependency.