From 64 items, 2 important content pieces were selected
Simon Willison observes that ‘vibe coding’ and agentic engineering are starting to overlap in his work, raising concerns about the software development lifecycle’s ability to handle AI-accelerated code production.
rss · Simon Willison · May 6, 14:24 · 社区讨论
标签: #AI coding, #software engineering, #vibe coding, #agentic engineering, #LLMs
sglang v0.5.11 upgrades to CUDA 13 and PyTorch 2.11, makes Speculative Decoding V2 default for lower CPU overhead, adds decode radix cache for prefill/decode disaggregation, and supports new models like Gemma 4 and Qwen3.6.
github · Kangyan-Zhou · May 5, 21:28
标签: #sglang, #release, #llm-serving, #performance-optimization, #speculative-decoding