2025年起,我反复强调的四个原则: 原则1,时代的第一性原理:LLM一定会越来越聪明,benchmark越来越高,context window越来越大,reasoning越来越长,价格越来越便宜,inference速度越来越快, 这是scaling law今天依然持续的具体方向,不用你质疑,这是你唯一的信仰和行业最大共识。 原则2, 管理学设计红利:从我提出“自动编程机”、行业提出vibe coding、SWE-Agent以来,从c…

开发者 lidang 立党在 X 上系统梳理了他从 2025 年起反复强调的四个 AI Agent 原则,核心判断是:LLM 能力持续变强是行业最大共识,但把 Agent 当人管、把管理学套在 multi-agent 工作流上的红利正在耗尽,Agent 真正的位置是嵌入生产环境的 runtime,而不是替代人…
![[Vulkan] GGML_ASSERT(neq0 == HSK) failed in ggml-vulkan.cpp during speculative draft decoding (MTP) with tensor split](https://www.chat-gpts.plus/wp-content/uploads/2026/10/29418-35687e0f-768x403.jpg)


![[Bug]: OpenAI-compatible Embeddings API intermittently crashes with multimodal cache assertion (`Expected a cached item for mm_hash`) on Qwe](https://www.chat-gpts.plus/wp-content/uploads/2026/10/33865-8d61767c-768x403.jpg)

![[bug]: Potential Same CUDA Graphs Kernel Path Failure Problem as Llama CPP? (Proposed workaround included)](https://www.chat-gpts.plus/wp-content/uploads/2026/10/9566-b918b8e9-768x403.jpg)

![[Bug]: llama-index-embeddings-azure-openai 0.6.0 is still dependent on llama-index-llms-azure-openai>=0.4.0,<0.6](https://www.chat-gpts.plus/wp-content/uploads/2026/10/23325-955c0d99-768x403.jpg)
