Briefing · 2026-07-28
技术简报 - 2026-07-28
On this page6 sections
The briefing's title, the editors' summary and their notes are written in Chinese, as are skill and knowledge entries.
Overview
#今天没有新信号：进来的五条里，两条是入门教程和职场研究稿，一条是三周前那则 Gemini 公告换了链接又被抓了一次，剩下两条没有正文可写。所以这期做的是补课——把三条一直没排上位置的信号放出来，它们恰好是「把模型变成能用的系统」的三层。Kimi K3 是权重那一层：2.8 万亿参数的旗舰，承诺月内开放权重，决定你有没有资格把模型握在自己手里。Gemini API 的托管智能体扩展是运行那一层：后台任务加远程 MCP 直连，意味着智能体不必再守着一个对话窗口活着。而 GitHub 那篇 Copilot 代码审查复盘是最容易被跳过、却最便宜的一层：他们把探索工具换成维护更好的通用 CLI，基准反而变差，真正把结果扳回来的是按「审查者实际怎么读 PR」重写指令——审查成本降了两成。三条连起来是同一句话：拿到权重、跑起来之后，决定成败的往往是最不像技术的那一层。
Immediate attention
#The highest-rated published signals; the editors' pinned items come first.
Kimi K3, and what we can still learn from the pelican benchmark
Moonshot AI announces Kimi K3, a 2.8-trillion-parameter flagship with an open-weight release promised by July 27 — the first 'open 3T-class' model...
For 负责模型选型与成本核算的团队、关注开放权重生态的研究者与自托管用户、评测与基准工程师
Expanding Managed Agents in Gemini API: background tasks, remote MCP and more
Google expands Managed Agents in the Gemini Interactions API: long-running background execution (start with background:true, poll/stream/reconnect by ID)...
For 构建生产级智能体的开发团队、评估 MCP 生态与工具接入方案的架构师、对比各家智能体平台能力的技术选型者
Worth tracking
#Worth following, but each needs more context or confirmation before it becomes a priority.
Better tools made Copilot code review worse. Here's how we actually improved it.
GitHub swapped Copilot code review's custom exploration tools for the better-maintained shared CLI tools (grep, glob...
For 在智能体框架上构建产品的工程团队、维护智能体工具层的平台工程师、负责智能体评测基准的工程师
Sources
#The public sources behind the published signals in this briefing.
Subscribe
The public feeds carry published briefings only, never drafts or archived ones. They are in Chinese.