Fti

vLLM's speculative decoding on AMD GPUs can verify multiple drafted tokens in a single pass, improving output-token throughput. Results vary by drafting method, model family, and workload.

Infrastructure • Hacker News

Productivity

Set up Meta Muse to delegate everyday tasks to an AI agent running on its own cloud computer.

Universal Music Group is launching an AI music platform powered by ElevenLabs, allowing users to create remixes and mashups using only properly licensed catalog tracks. The partnership establishes a formalized revenue-sharing and licensing model for generative audio, potentially shifting industry standards away from unauthorized scraping.

Media & Entertainment • The Verge AI