X 时间线过去24小时：Kimi K3、Grok 与推理硬件
本期全量覆盖过去24小时的 X 关注时间线，聚合 Kimi K3、Grok/Claude 编码价值、推理硬件、Agent/Skills 工作流和开源模型商业影响。
本期覆盖 2026-07-17 09:00 至 2026-07-18 09:00（Asia/Shanghai）的 X 关注时间线：400 个关注账号全量抓取成功，8 批抓取失败账号为 0，窗口内筛出 511 条候选。
今天的主线不是单条推文冲高，而是开源模型继续把编码能力、推理硬件和商业价值分配拉到同一张桌上。Kimi K3 的讨论从产品演示扩展到 2.8T 体量和推理硬件约束；Grok 与 Claude 的竞争落在速度、成本、编码价值和开发者实测；同时，Agent / Skills 入口、快速推理硬件、开放权重商业影响，也成为关注圈里的高信号分支。
片尾推荐关注：@SemiAnalysis_、@hardmaru、@nikitabier、@AravSrinivas、@steipete。
来源
- @rsalakhu：I’ve been asked several times whether Zhilin Yang, the founder of @Kimi_Moonshot was m
- @levelsio：MacOS 27 in the browser created by Kimi K3 https://t.co/gzxHGPdIx2
- @SemiAnalysis_：Kimi K3 2.8T is so large that it will not fit on a single NVIDIA DGX B200, even at FP4
- @elonmusk：Grok 4.5 is arguably #1 when taking speed and cost into account
- @grok：Grok Build and Grok 4.5 top user model preference
- @nikitabier：If you want to keep track of the latest AI model releases, pin the AI custom timeline.
- @AndrewYNg：New course: Build LLM applications that respond to user requests quickly by running on
- @hardmaru：Real brains follow Dale's principle: a neuron can either excite its neighbors or suppr
- @AravSrinivas：At its peak, Sun Microsystems was valued at 205B (394B if inflation adjusted). Sold so
- @emollick：I am confused about the belief that if open weights eventually dominate it will lead t
- @steipete：5.6 Terra high is underrated. Switched @clawsweeper (GitHub review bot) to it and it's
- @rileybrown：Codex Skills don't transfer to GPT Work. This was / is one of the most annoying parts
- @swyx：btw if you havent set your {codex | claude | gemini | devin} automations to autoresear
相似内容
- 登录后可发表评论。