@victormustar@huggingface Do you think MTP will meaningfully improve thermally constrained devices like Android phones, or is memory bandwidth still the main bottleneck?
Been experimenting with local LLM inference on Android lately
Honestly one of the most interesting parts has been learning how hard mobile inference tuning actually is.
Surprisingly, CPU-only TinyLlama sometimes outperformed Vulkan hybrid inference on my device.
#gemma4#llama
The 2026 Stanford AI Index just dropped. AI models went from 8.8% to 50%+ on PhD-level questions in a single year. This is the best time in history to be building with AI. We are so early.
#stanford#ai
Anthropic built an AI so powerful they refused to release it. Claude Mythos: 94.6% on GPQA Diamond, found thousands of zero-days on its own. Too dangerous for the public. We are NOT in normal times anymore. What does that tell you about where this is heading?
dedicated to staying consistent to level-up my DSA knowledge and skills , implementing structured learning strategies and engaging with resources to continuously improve my problem solving skills. #DSA#ezsnippet