332 lines
31 KiB
XML
332 lines
31 KiB
XML
<?xml version="1.0" encoding="utf-8"?>
|
||
<feed xmlns="http://www.w3.org/2005/Atom">
|
||
<title>Spring 编辑精选</title>
|
||
<id>tag:mzaxd,2026:newsroom</id>
|
||
<updated>2026-09-03T23:34:39+08:00</updated>
|
||
<subtitle>nanobot 每周二/五从知乎/Reddit 等平台人工筛选，按你的画像定制</subtitle>
|
||
<entry>
|
||
<title>ChatGPT、Grok、Claude、Cursor 集体突发故障：主流 AI 服务同时宕机</title>
|
||
<id>tag:mzaxd,2026-09-03:n1</id>
|
||
<link href="https://www.zhihu.com/question/2078984478073136308"/>
|
||
<updated>2026-09-03T23:34:39+08:00</updated>
|
||
<published>2026-09-03T23:34:39+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>四家头部 AI 服务 9 月 3 日晚同时不可用——若根因落在共享依赖（云厂商/推理供应商层），就是 AI 基础设施单点风险的现场教学。多供应商冗余（你的 CPA 网关思路）在事故夜的价值会被重新讨论，等各家 post-mortem 出来对照。</p><p><a href="https://www.zhihu.com/question/2078984478073136308">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>字节跳动将获约 296 亿美元银团贷款，全年最高 700 亿美元 AI 资本开支</title>
|
||
<id>tag:mzaxd,2026-09-03:n2</id>
|
||
<link href="https://www.zhihu.com/question/2078880121335952288"/>
|
||
<updated>2026-09-03T23:34:39+08:00</updated>
|
||
<published>2026-09-03T23:34:39+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>700 亿美元年资本开支已对齐美国超大规模云厂商量级，「字节正在变成基建公司」的讨论本质是：模型军备竞赛的入场券价格被抬到只有巨头能付。你跟踪港股 AI 资产定价，中国 AI 大厂的资本强度曲线是估值模型的关键输入。</p><p><a href="https://www.zhihu.com/question/2078880121335952288">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>官方确认：Nvidia 以 129 亿美元收购 Hugging Face</title>
|
||
<id>tag:mzaxd,2026-09-03:n3</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1w65uhf/its_official_nvidia_to_acquire_hugging_face_for/"/>
|
||
<updated>2026-09-03T23:34:39+08:00</updated>
|
||
<published>2026-09-03T23:34:39+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>上期收录的「收购谈判中」现已官宣，129 亿美元落定。开源社区担忧的点不在平台关停而在中立性：GPU 厂商坐拥模型分发入口后的利益冲突，以及 llama.cpp 团队随之易主的下游影响。开源基础设施被硬件巨头收编的进程正式完成。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1w65uhf/its_official_nvidia_to_acquire_hugging_face_for/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>Qwen 3.8 27B 在 16GB 显卡上跑出 50 tok/s + 100k 上下文（beellama.cpp）</title>
|
||
<id>tag:mzaxd,2026-09-03:n4</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1w1lq7u/qwen_38_27b_at_50_toks_with_100k_context_on_a/"/>
|
||
<updated>2026-09-03T23:34:39+08:00</updated>
|
||
<published>2026-09-03T23:34:39+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>消费级显卡跑本地大模型的效率天花板又被抬高：27B 级模型 16GB 显存下 50 tok/s、百 k 级上下文。对你「等 24G 卡重启本地 LLM 线」的计划是利好信号——软件侧（量化/内核/上下文管理）的进步速度不输硬件，届时可用门槛只会更低。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1w1lq7u/qwen_38_27b_at_50_toks_with_100k_context_on_a/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>腾讯把 Hy4-preview 从 1.5TB 压到约 200GB GGUF，性能保持 ~98%</title>
|
||
<id>tag:mzaxd,2026-09-03:n5</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1w1o324/tencent_compressed_hy4preview_from_15tb_to_about/"/>
|
||
<updated>2026-09-03T23:34:39+08:00</updated>
|
||
<published>2026-09-03T23:34:39+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>770B-A49B 的 Hunyuan4 官方放权后社区跟进的压缩成果：体积砍到 1/7、性能保留 98%，大 MoE 模型「跑得动」的边界被大幅外推。与 Qwen3.8-Flash-Next 的 n-gram 路线相互印证：端侧跑旗舰模型的工程路径正在收敛。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1w1o324/tencent_compressed_hy4preview_from_15tb_to_about/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>Terminal Bench 4.0 出榜：GLM-5.3 与 Fable 5 同档</title>
|
||
<id>tag:mzaxd,2026-09-03:n6</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1w1fpxi/terminal_bench_40_just_dropped_glm53_is_at_the/"/>
|
||
<updated>2026-09-03T23:34:39+08:00</updated>
|
||
<published>2026-09-03T23:34:39+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>agentic coding 基准 Terminal Bench 4.0 放榜，GLM-5.3 进入第一梯队。你手里的 GLM Coding Plan（¥1164/年，年底评估续费）又添一个第三方实证——模型能力持续在线是续费评估的正面项，计划到期前这份榜单值得翻一遍。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1w1fpxi/terminal_bench_40_just_dropped_glm53_is_at_the/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>内存、存储、显卡集体涨价：等等党要等到 2040 年吗？</title>
|
||
<id>tag:mzaxd,2026-09-03:n7</id>
|
||
<link href="https://www.zhihu.com/question/2078521113076830363"/>
|
||
<updated>2026-09-03T23:34:39+08:00</updated>
|
||
<published>2026-09-03T23:34:39+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>AI 需求把 DRAM/NAND 产能吃穿的传导开始波及消费端——显卡、内存、SSD 全线上涨。你名下挂着两个硬件计划（RTX 5070 Ti Super 等 CES 2027、NAS 扩容盘），「按需即买、不做等等党」在这轮周期里大概率是对的，扩容计划别因涨价无限期推迟。</p><p><a href="https://www.zhihu.com/question/2078521113076830363">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>Let's Encrypt 证书有效期缩短在即：homelab 该做什么准备</title>
|
||
<id>tag:mzaxd,2026-09-03:n8</id>
|
||
<link href="https://www.reddit.com/r/selfhosted/comments/1w49uhi/how_i_chose_to_ready_my_homelab_for_the_upcoming/"/>
|
||
<updated>2026-09-03T23:34:39+08:00</updated>
|
||
<published>2026-09-03T23:34:39+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>证书有效期向 45-47 天压缩的滚动计划已在路上，手动续期的玩法正式出局，ACME 全自动化成为硬门槛。你的 NPM 泛域名 + DNS challenge 本身是自动续期，但值得找时间确认一遍 cpa.mzaxd.fun 等非 NPM 管辖端点的续期链路没有掉队。</p><p><a href="https://www.reddit.com/r/selfhosted/comments/1w49uhi/how_i_chose_to_ready_my_homelab_for_the_upcoming/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>知乎热议：OpenAI Codex 将取消上下文压缩，改用「硬切窗口 + 外部记忆」</title>
|
||
<id>tag:mzaxd,2026-08-31:n1</id>
|
||
<link href="https://www.zhihu.com/question/2081388573025586179"/>
|
||
<updated>2026-08-31T23:31:57+08:00</updated>
|
||
<published>2026-08-31T23:31:57+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>coding agent 的上下文管理路线之争浮出水面：自动压缩（summarize 续写）被砍掉，改走硬切窗口+外部记忆检索。压缩保连贯但失真累积，硬切+检索保真实但丢脉络——两条路线的取舍会直接影响长任务 agent 的工程形态。你重度使用 coding agent 且 nanobot 本身就是 1M ctx 重放式设计，这题值得一看。</p><p><a href="https://www.zhihu.com/question/2081388573025586179">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>宇树股价「腰斩」：1100 跌至 555 元，2200 亿市值蒸发</title>
|
||
<id>tag:mzaxd,2026-08-31:n2</id>
|
||
<link href="https://www.zhihu.com/question/2081390608017122297"/>
|
||
<updated>2026-08-31T23:31:57+08:00</updated>
|
||
<published>2026-08-31T23:31:57+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>对照你 8 月的估值判断（发行价 150.8、上市初期 672 属筹码稀缺+情绪溢价、合理中枢 150-200）——股价冲到 1100 后腰斩至 555，验证情绪出清逻辑，但仍高于你给的中枢区间，A 股无做空工具背景下回归节奏只会更慢。机器人第一股的估值锚正在被市场重新定价。</p><p><a href="https://www.zhihu.com/question/2081390608017122297">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>知乎热议：Kimi K3 的实际表现真有被吹捧的那么强吗？</title>
|
||
<id>tag:mzaxd,2026-08-31:n3</id>
|
||
<link href="https://www.zhihu.com/question/2081389268592241751"/>
|
||
<updated>2026-08-31T23:31:57+08:00</updated>
|
||
<published>2026-08-31T23:31:57+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>国产旗舰模型发布后的口碑周期管理题：发布日 benchmark 惊艳与日常体验落差的老话题。Reddit 侧也在等 Kimi K3 的编码表现（与 Qwen3.8 27B 消费级神话对照看），质疑向讨论恰好提供冷静视角。</p><p><a href="https://www.zhihu.com/question/2081389268592241751">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>国内首部 AIGC 长剧《后西游记》开播，登陆湖南卫视黄金档</title>
|
||
<id>tag:mzaxd,2026-08-31:n4</id>
|
||
<link href="https://www.zhihu.com/question/2081390004256391045"/>
|
||
<updated>2026-08-31T23:31:57+08:00</updated>
|
||
<published>2026-08-31T23:31:57+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>AI 视频赛道从工具走向内容产品的标志性事件：AIGC 长剧首次进入主流卫视黄金档。制作成本结构与产能逻辑和传统影视完全不同，效果口碑如何将直接影响资本对 AI 视频内容化的预期——你跟踪可灵这类赛道，这是需求侧的实证样本。</p><p><a href="https://www.zhihu.com/question/2081390004256391045">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>英伟达 129 亿美元收购 Hugging Face：开源基础设施被 GPU 巨头收编</title>
|
||
<id>tag:mzaxd,2026-08-27:n1</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1vzfwnd/nvidia_has_been_in_talks_to_acquire_hugging_face/"/>
|
||
<updated>2026-08-27T23:33:31+08:00</updated>
|
||
<published>2026-08-27T23:33:31+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>Business Insider 报道收购价超 130 亿美元，知乎热榜同题讨论已起。连带效应更值得看：llama.cpp 团队随之易主，社区担忧开源中立性。HF 是整个开源模型世界的分发层，被英伟达拿下的战略含义远超交易金额本身——模型生态的上下游话语权正在集中。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1vzfwnd/nvidia_has_been_in_talks_to_acquire_hugging_face/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>知乎热议：Anthropic 最强模型 Fable 5 遇冷，「更强模型」被高估了吗？</title>
|
||
<id>tag:mzaxd,2026-08-27:n2</id>
|
||
<link href="https://www.zhihu.com/question/2075227159929386027"/>
|
||
<updated>2026-08-27T23:33:31+08:00</updated>
|
||
<published>2026-08-27T23:33:31+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>与你周三收的 Claude Code 退订潮同一条主线：旗舰模型发布遇冷 + 用户流向竞品，行业开始重估'更强模型'的商业价值。支撑你的核心判断——AI 盈利困境的瓶颈不在模型智力上限，而在算力成本与需求错配。</p><p><a href="https://www.zhihu.com/question/2075227159929386027">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>智谱发布 GLM-5.3-Flash：Frontier 智能，Flash 成本</title>
|
||
<id>tag:mzaxd,2026-08-27:n3</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1vyy3k6/glm53flash_frontier_intelligence_flash_cost/"/>
|
||
<updated>2026-08-27T23:33:31+08:00</updated>
|
||
<published>2026-08-27T23:33:31+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>智谱把旗舰能力下探到 Flash 档定价。你 nanobot 主力跑的就是 glm-5.3（GLM Coding Plan 年底评估续费），Flash 档出现直接影响你的成本结构——如果价格进一步下探，'弃订需先改抽取模型'的成本顾虑也会松动。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1vyy3k6/glm53flash_frontier_intelligence_flash_cost/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>Qwen3.8-Flash-Next：~125B-A6B + 51B n-gram，本地部署新架构路线</title>
|
||
<id>tag:mzaxd,2026-08-27:n4</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1vy6smx/qwen38flashnext_this_architecture_could_be/"/>
|
||
<updated>2026-08-27T23:33:31+08:00</updated>
|
||
<published>2026-08-27T23:33:31+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>MoE 125B 总参/6B 激活，外挂 51B n-gram 表——Unsloth 已 day-0 出量化（4-bit 理想约 82GB）。另有澄清帖专门解释 Engrams/n-gram 表并不会让人本地跑 1T 模型，把预期管理讲清楚了。大厂开始为'本地友好'专门设计架构，这条产品线值得跟踪。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1vy6smx/qwen38flashnext_this_architecture_could_be/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>苹果 Mac Studio M5 Ultra 正式发布：512GB 统一内存、1.2TB/s 带宽</title>
|
||
<id>tag:mzaxd,2026-08-27:n5</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1vxzg6v/apple_introduces_new_mac_studio_with_m5_max_and/"/>
|
||
<updated>2026-08-27T23:33:31+08:00</updated>
|
||
<published>2026-08-27T23:33:31+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>上周还是传闻，本周落地：M5 Ultra 1.2TB/s 带宽对齐小米 AI Cube，512GB 统一内存意味着 400B 级模型本地可跑。消费级推理硬件军备竞赛成型（Apple/小米/英伟达 DGX Spark），GPU+HBM 之外，统一内存路线正在分走'本地大模型'的心智。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1vxzg6v/apple_introduces_new_mac_studio_with_m5_max_and/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>PSA：避免自托管 AppFlowy——SaaS 侧修了漏洞，自托管版被晾着</title>
|
||
<id>tag:mzaxd,2026-08-27:n6</id>
|
||
<link href="https://www.reddit.com/r/selfhosted/comments/1vyk94h/psa_avoid_self_hosting_appflowy_they_patched_a/"/>
|
||
<updated>2026-08-27T23:33:31+08:00</updated>
|
||
<published>2026-08-27T23:33:31+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>开源+商业 SaaS 双轨项目的经典信任问题：安全补丁先到 SaaS，自托管用户是二等公民。教训通用化——选自托管项目时，看它对自托管社区的态度比看 star 数重要；你拒绝低信任项目的直觉在这类事件里反复被验证。</p><p><a href="https://www.reddit.com/r/selfhosted/comments/1vyk94h/psa_avoid_self_hosting_appflowy_they_patched_a/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>ExperiencedDevs：如何在巨型项目里评估改动的爆炸半径</title>
|
||
<id>tag:mzaxd,2026-08-27:n7</id>
|
||
<link href="https://www.reddit.com/r/ExperiencedDevs/comments/1vxz9g9/how_do_you_estimate_blast_radius_impact_of/"/>
|
||
<updated>2026-08-27T23:33:31+08:00</updated>
|
||
<published>2026-08-27T23:33:31+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>接手巨型遗留系统后如何评估一次变更的影响面——评论区从依赖图、灰度策略到'哪些地方没人敢动'的 tribal knowledge 挖掘都有。对你在天然气平台这类多乙方长周期项目里有直接参考价值。</p><p><a href="https://www.reddit.com/r/ExperiencedDevs/comments/1vxz9g9/how_do_you_estimate_blast_radius_impact_of/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>知乎热议：Claude Code 增长熄火、传大规模退订，老用户转投 Codex</title>
|
||
<id>tag:mzaxd,2026-08-24:n1</id>
|
||
<link href="https://www.zhihu.com/question/2074071189530440759"/>
|
||
<updated>2026-08-24T23:33:35+08:00</updated>
|
||
<published>2026-08-24T23:33:35+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>你每天在用的主力工具正处在舆论拐点：定价/配额争议 + Codex 抢用户。你手里 CPA 网关同时接着 Claude Code 和 Codex，正好用高赞回答里的真实吐槽对照自己的使用体验，判断这波是情绪还是产品力真变了。</p><p><a href="https://www.zhihu.com/question/2074071189530440759">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>Agent 目前最大的瓶颈是什么？（知乎老题翻红）</title>
|
||
<id>tag:mzaxd,2026-08-24:n2</id>
|
||
<link href="https://www.zhihu.com/question/1986872411988173762"/>
|
||
<updated>2026-08-24T23:33:35+08:00</updated>
|
||
<published>2026-08-24T23:33:35+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>2023 年的老问题本周重新冲榜，说明 agent 落地讨论进入新阶段。你跟踪 OpenClaw/nanobot 这层生态，高赞回答对长任务可靠性、上下文管理、工具调用闭环失败的归因值得扫一遍——多数痛点与你自己的 agent 运维经验可以互证。</p><p><a href="https://www.zhihu.com/question/1986872411988173762">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>Qwen3.8-27B agentic coding 证据链升级：接进 Codex 对比 GPT Luna、Q6 量化实战、39k 行 C 移植</title>
|
||
<id>tag:mzaxd,2026-08-24:n3</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1vvyacg/qwen_38_27b_is_a_game_changer/"/>
|
||
<updated>2026-08-24T23:33:35+08:00</updated>
|
||
<published>2026-08-24T23:33:35+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>上周是基准刷屏，本周是干活实录三连：团队把它接进 Codex 与 GPT Luna 对比日常开发任务；Q6 量化版 agentic coding 表现惊艳；还有人用它把 39k 行 C 代码移植成单文件 three.js。27B 级模型开始真正替代云端干活，Q4/Q6 量化对 10G 显存也亲民。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1vvyacg/qwen_38_27b_is_a_game_changer/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>小米 AI Cube 原型机：1.2TB/s 内存带宽的三芯片系统</title>
|
||
<id>tag:mzaxd,2026-08-24:n4</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1vwvghi/xiaomi_ai_cube_announced_with_12tbs_memory/"/>
|
||
<updated>2026-08-24T23:33:35+08:00</updated>
|
||
<published>2026-08-24T23:33:35+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>玄戒 O3 等三芯片架构，1.2TB/s 带宽直接对标 Mac Studio 级推理硬件。消费大厂集体下场本地推理，叠加内存涨价 500% 的背景，'算力不够便宜'这个你反复强调的行业瓶颈，供给端正在起变化。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1vwvghi/xiaomi_ai_cube_announced_with_12tbs_memory/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>阿里玄铁 C950：RISC-V CPU 跑 Qwen-3.8 27B 到 30 tps</title>
|
||
<id>tag:mzaxd,2026-08-24:n5</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1vs0wsl/alibabas_riscv_cpu_xuantie_c950_runs_qwen38_27b/"/>
|
||
<updated>2026-08-24T23:33:35+08:00</updated>
|
||
<published>2026-08-24T23:33:35+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>不用 GPU、纯 RISC-V CPU 跑 27B 到 30 tps。开源模型 + 开源指令集的组合是推理硬件多元化的最新样本——GPU/HBM 被内存危机卡脖子时，CPU+大内存路线的可行性边界在哪，这帖是个数据点。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1vs0wsl/alibabas_riscv_cpu_xuantie_c950_runs_qwen38_27b/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>智谱研究院长文《Thoughts About Scaling Law》</title>
|
||
<id>tag:mzaxd,2026-08-24:n6</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1vsf9eg/thoughts_about_scaling_law_zai/"/>
|
||
<updated>2026-08-24T23:33:35+08:00</updated>
|
||
<published>2026-08-24T23:33:35+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>智谱对 scaling law 的系统性反思：不只是参数量，数据和 agent RL 同样是可缩放维度。你在做 MiniMax vs 智谱的估值对比，这篇等于智谱对行业路线图的公开表态，判断其技术叙事的一手材料。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1vsf9eg/thoughts_about_scaling_law_zai/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>Meelo v3.12.0：主打 UI 与元数据的自托管音乐服务器</title>
|
||
<id>tag:mzaxd,2026-08-24:n7</id>
|
||
<link href="https://www.reddit.com/r/selfhosted/comments/1vsoqno/meelo_v3120_music_server_focused_on_ui_metadata/"/>
|
||
<updated>2026-08-24T23:33:35+08:00</updated>
|
||
<published>2026-08-24T23:33:35+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>与 Navidrome 差异化竞争的路线：元数据管理和 UI 优先。你库里 2 万曲目 genre 标签基本为空、做不了曲风歌单的痛点，可以看它怎么解——不一定换库，但它处理元数据的产品思路值得偷师。</p><p><a href="https://www.reddit.com/r/selfhosted/comments/1vsoqno/meelo_v3120_music_server_focused_on_ui_metadata/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>「还有人用裸 Linux + Docker Compose 跑 homelab 吗？」——社区大讨论</title>
|
||
<id>tag:mzaxd,2026-08-24:n8</id>
|
||
<link href="https://www.reddit.com/r/selfhosted/comments/1vw1oqz/does_anyone_still_run_their_homelab_on_plain/"/>
|
||
<updated>2026-08-24T23:33:35+08:00</updated>
|
||
<published>2026-08-24T23:33:35+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>与你的 Unraid dockerMan 模板标准化路线正反对照：评论区大量裸机 vs 发行版的真实取舍。你禁掉 Compose Manager 随手起栈、坚持模板化管理的决策，在这个帖子里能同时找到支持者和反例。</p><p><a href="https://www.reddit.com/r/selfhosted/comments/1vw1oqz/does_anyone_still_run_their_homelab_on_plain/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>高年资工程师如何持续学习而不把它变成第二份工作</title>
|
||
<id>tag:mzaxd,2026-08-24:n9</id>
|
||
<link href="https://www.reddit.com/r/ExperiencedDevs/comments/1vslfiy/strategies_to_keep_learning_without_tuning_it/"/>
|
||
<updated>2026-08-24T23:33:35+08:00</updated>
|
||
<published>2026-08-24T23:33:35+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>ExperiencedDevs 本周质量最高帖之一：不是鸡汤，是具体的学习系统设计。与你上月诊断出的'输入过剩、内化断链、输出半成品'问题完全同构，评论区的一手习惯设计可以直接参考。</p><p><a href="https://www.reddit.com/r/ExperiencedDevs/comments/1vslfiy/strategies_to_keep_learning_without_tuning_it/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>Stripe 拟 70 亿美元收购 AI 网关 OpenRouter，路由层开始整合</title>
|
||
<id>tag:mzaxd,2026-08-22:n1</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1vqlh98/stripe_will_reportedly_acquire_ai_gateway_startup/"/>
|
||
<updated>2026-08-22T17:30:00+08:00</updated>
|
||
<published>2026-08-22T17:30:00+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>AI 网关（gateway/router）层出现首笔巨型并购。对你的直接启示：CPA 这类自建网关的价值判断需要重新校准——支付巨头下场说明路由层被认为是 AI 商业基础设施的咽喉。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1vqlh98/stripe_will_reportedly_acquire_ai_gateway_startup/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>彭博：美国 AI 优势正被中国迅速缩小（知乎热榜讨论）</title>
|
||
<id>tag:mzaxd,2026-08-22:n2</id>
|
||
<link href="https://www.zhihu.com/question/2074086385397544896"/>
|
||
<updated>2026-08-22T17:30:00+08:00</updated>
|
||
<published>2026-08-22T17:30:00+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>彭博视角的中美 AI 差距收敛叙事。与你跟踪的 MiniMax/智谱估值判断直接相关：如果差距收敛论成立，中国 AI 资产的折价逻辑会松动。知乎高赞回答里有不少模型层面的具体对比。</p><p><a href="https://www.zhihu.com/question/2074086385397544896">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>DeepSeek V4 Flash Vision Exp 发布（0821）</title>
|
||
<id>tag:mzaxd,2026-08-22:n3</id>
|
||
<link href="https://www.zhihu.com/question/2074187767911925314"/>
|
||
<updated>2026-08-22T17:30:00+08:00</updated>
|
||
<published>2026-08-22T17:30:00+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>DeepSeek 给 V4 Flash 加了视觉能力（实验版）。知乎当天就有评测讨论。值得关注它在本地部署场景的显存表现——3080 10G 能不能跑是个现实问题。</p><p><a href="https://www.zhihu.com/question/2074187767911925314">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>Qwen3.8-27B 本周刷屏：Unsloth Dynamic v3 量化 + 基准对齐 DeepSeek</title>
|
||
<id>tag:mzaxd,2026-08-22:n4</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1vqyq8r/artificial_analysis_qwen3827b_benchmarks_put_it/"/>
|
||
<updated>2026-08-22T17:30:00+08:00</updated>
|
||
<published>2026-08-22T17:30:00+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>Qwen3.8-27B 是本周 LocalLLaMA 主旋律：Artificial Analysis 基准把它排到与 DeepSeek 同档，Unsloth 的 Dynamic v3 量化同尺寸精度提升 10%。另有帖子称它是'本地模型里 agency 最强'——对你折腾本地 agent 有直接参考价值。量化版：reddit.com/r/LocalLLaMA/comments/1vsr67c</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1vqyq8r/artificial_analysis_qwen3827b_benchmarks_put_it/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>内存价格 12 个月上涨 500%</title>
|
||
<id>tag:mzaxd,2026-08-22:n5</id>
|
||
<link href="https://www.reddit.com/r/LocalLLaMA/comments/1vrwsfl/memory_prices_climb_500_in_12_months_up_to_10x/"/>
|
||
<updated>2026-08-22T17:30:00+08:00</updated>
|
||
<published>2026-08-22T17:30:00+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>DRAM 涨价已到历史级，AI 数据中心需求抽干了消费级供给。短期信号：想加内存条/换 NAS 内存趁早，2026 年内看不到缓解。</p><p><a href="https://www.reddit.com/r/LocalLLaMA/comments/1vrwsfl/memory_prices_climb_500_in_12_months_up_to_10x/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>SparkyFitness v1.6.2：自托管 MyFitnessPal 替代（5.2k stars）</title>
|
||
<id>tag:mzaxd,2026-08-22:n6</id>
|
||
<link href="https://www.reddit.com/r/selfhosted/comments/1vqe7qg/sparkyfitness_v162_a_selfhosted_alternative_for/"/>
|
||
<updated>2026-08-22T17:30:00+08:00</updated>
|
||
<published>2026-08-22T17:30:00+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>开源饮食/健康追踪，98 人贡献。与你的 AI 饮食健康 App 方向直接对标——建议拉下来看看它的数据模型和 AI 功能边界，竞品调研成本几乎为零。</p><p><a href="https://www.reddit.com/r/selfhosted/comments/1vqe7qg/sparkyfitness_v162_a_selfhosted_alternative_for/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>「用 AI 写代码之后，这个职业到底在往哪走？」——ExperiencedDevs 本周高热讨论</title>
|
||
<id>tag:mzaxd,2026-08-22:n7</id>
|
||
<link href="https://www.reddit.com/r/ExperiencedDevs/comments/1vq8w8c/whats_wrong_with_our_team/"/>
|
||
<updated>2026-08-22T17:30:00+08:00</updated>
|
||
<published>2026-08-22T17:30:00+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>7 年经验工程师的真实困惑：AI 写码后个人价值锚点在哪。评论区有大量一线工程师的应对策略。与《The Pragmatic Engineer》近期 Meta 裁员潮报道互为印证——工程职业结构正在换挡。</p><p><a href="https://www.reddit.com/r/ExperiencedDevs/comments/1vq8w8c/whats_wrong_with_our_team/">阅读原文</a></p></content>
|
||
</entry>
|
||
<entry>
|
||
<title>非科技公司升 Staff 工程师 8 个月复盘</title>
|
||
<id>tag:mzaxd,2026-08-22:n8</id>
|
||
<link href="https://www.reddit.com/r/ExperiencedDevs/comments/1vpg59k/followup_what_this_year_has_been_like_as_a/"/>
|
||
<updated>2026-08-22T17:30:00+08:00</updated>
|
||
<published>2026-08-22T17:30:00+08:00</published>
|
||
<author><name>nanobot</name></author>
|
||
<content type="html"><p>大型非科技公司（接近你司环境）从 Senior 升 Staff 的完整复盘：晋升后的期望错位、技术影响力如何落地。对判断'走管理还是走专家'有第一手参照。</p><p><a href="https://www.reddit.com/r/ExperiencedDevs/comments/1vpg59k/followup_what_this_year_has_been_like_as_a/">阅读原文</a></p></content>
|
||
</entry>
|
||
</feed>
|