Centralising entitlement activation turns a per-account chore into an automated pipeline, which matters once model pricing agreements span a multi-account organisation. Platform and FinOps teams managing Bedrock at scale can adopt the CDK pattern to handle new entitlements automatically. The script option covers teams that only need a one-off cleanup.
NVIDIA 开了一场直播，手把手教你怎么微调 ASR 和 LLM，把通用语音模型改成某个行业的专用语音智能体，做语音方向的可以看看。
Bolt 放出一条现成提示词，直接复制就能生成带地图的门店定位页，用 OpenFreeMap 连 API key 都不用申请，做展示页很省事。
Garry Tan 推荐了 AsideAI，能让编程智能体直接用你登录好的浏览器操作网页，比单纯调 API 灵活多了。
Anthropic 把内部工程博客公开了：合并量涨十倍后他们干脆自研版本控制，五个月迁移完，做法和坑都写得挺细。
作者聊了自己同时用 ChatGPT Dots、Grokbot、Muse 三个 Bot 的真实感受，为啥有的留得住、有的留不住，挺接地气。
Fish Audio 的 Drama 3 可以用大白话控制语音语气，作者演示了怎么把论文自动变成播客，流程代码都给了。
ShipStation 把 IT 支持做成 Slack 里的智能体，三个月解决了 97.5% 的提问，还带截图识别和自动开 Jira 工单，搭建思路值得参考。
Reasoning effort is a per-request dial, so teams can spend more tokens only on hard steps and keep routine tool calls cheap, trading accuracy against cost with real numbers as reference. Developers building agents on Bedrock get a working evaluation pattern they can copy for cost control in production. Availability on Bedrock also means access without a direct xAI contract.
用 Vercel AI SDK 的开发者注意，官方文档里加了 Google 模型的接入说明，照着做就能跑通调用。
The implication is that inference infrastructure built and benchmarked for chat traffic can misprice agent serving, where cache hit rate on repeated prefixes dominates cost and latency. Engineers running agentic backends should evaluate serving stacks by prefix cache hit rate and KV reuse rather than headline throughput. The session also signals vLLM is being pushed as a production option for agent serving, against managed offerings from hyperscale clouds.
Google 晒了三个 Argon 实战案例：重构 Chrome 600 万行代码、跑 ADK 省 40% 准备时间、Gemma 4 推理提速 2.85 倍，做法都能抄。
一条把 vibe coding 玩到极致的段子：财务报告、聊天应用、PDF 阅读器、操作系统全是生成的，看完会心一笑。
The benchmark suggests a practical routing rule: swap general-purpose LLM calls for Jev only on narrow, high-frequency decisions like stopping checks, and keep dedicated rerankers like Voyage rerank-3 for re-ranking. Teams running RAG pipelines can use the latency and cost figures to decide where Jev cuts spend without hurting recall. Full results are on Milvus's blog.
The cost-comparison section gives practitioners a concrete framework for choosing between self-hosted GPUs, subscriptions and token-based API billing, a decision most teams still make ad hoc. Builders running LLM endpoints in production will get actionable patterns for autoscaling, batching and MoE serving on Ray Serve. Registration is open via the link in the announcement.
这位博主攒了 5 个 Claude skill 没发，字幕剪辑、公众号排版、HTML 排版都有，正在投票看先放哪个，感兴趣可以去投。
歸藏分享的 GrokBot 用法，配一个 HTML 排版 skill，能把推串和长视频转成手机上好读的中文长文，刷到干货推串时可以直接这么处理。
Weaviate 这篇教你按自己的数据选 embedding 模型，还拿 OpenAI 大小模型实测对比，方法可直接照抄。
The techniques target the exact bottlenecks that slow self-hosted MoE inference, and they are documented well enough to reproduce in your own SageMaker deployments. Teams serving large sparse models on NVIDIA H200 instances can apply them directly to cut latency and cost per token. Compared with relying on a hosted model API, this matters most for builders who need control over quantization and serving internals.
想搭靠谱 Agent 的朋友看这个：一篇讲清模型、数据、安全、基础设施那些坑的峰会介绍，NVIDIA 和 Hugging Face 的人都来。
做 AI 套壳应用的都该看看，你给 Claude 或 GPT 传的每一条数据，都在帮上游学会怎么替代你。
Naresh Jain 实测发现给前沿模型套 harness 反而多花 token，弱模型却能明显变强，省钱思路很实用。
mondaydotcom 的工程师讲怎么用 LangSmith 搭评测管线，线上监控加线下回归都有，做 LLM 应用的可以抄作业。
用 Codex CLI 的朋友注意了，新版会偷偷常驻后台，自动更新还会打断长任务，文中有个命令可以关掉。
教你把一个 ChatGPT 或 Claude 会话当总控台，去调度你开的其他会话，连建新会话、改名、分组都能管，很实用的用法。
作者聊了个观察：各公司都在找能把数据和 Agent 串起来的工具，Glass 思路不错但还没实测，适合想引入 Agent 协作的人参考。
svpino 用会议改期这个例子讲清了 flat 检索为什么不行，Oracle AI Agent Memory 26.8 的 graph-aware 方案有代码可跑，做 RAG 的可以试试。
Instructions in AGENTS.md are advisory; a git pre-commit hook turns them into a hard gate the agent cannot bypass, closing the gap between documented and enforced conventions. Developers running Claude Code, Cursor or similar coding agents can adapt this pattern to any team rule they want enforced automatically. It is a concrete alternative to prompt-based guardrails that quietly degrade over time.
有人用 Claude 和 Codex 做逆向翻车了，提醒大家别拿主力账号去跑这类任务，逆向代码也别随便发 GitHub。
有人给 Obsidian 写了个小 Agent，说一句话就能自动整理 inbox 里的笔记，重度笔记用户可以抄作业。
三场实战演讲打包看：OVO Energy 怎么做安全客服智能体，还有 LangSmith Engine 内部构建思路，做智能体的能抄作业。
印尼快递公司 Lion Parcel 拿 Qwen-VL-Plus 自动识别运单，不用再人工敲数据，做物流或有大量单据的可以参考。
LangChain 官方出学习课程了，想上手 Agent 开发的可以直接跟着官方教程练，不用自己东拼西凑找资料。
Only the latest 2000 items are listed; for older ones see the daily archive
