Topic

Coding agents

Agents that write, review and change code, and what they change in engineering practice.

Follow this topic

On this site · Elsewhere
On this site · My radar
Elsewhere · RSS reader

Topic graph

16 nodes · 58 links
The graph shows the 10 of the 24 entries under “Coding agents” that are most connected to each other, and the 6 of the topic's 70 direct outside neighbours most connected to them. The full lists are below.
  • Entries in this topic
  • Neighbours outside the topic
Relation types
  • Builds on · Successor · Extends
  • Prerequisite · Uses
  • Explains · Supports
View in the network →

Skills and knowledge entries are written in Chinese, as are signals from Chinese sources.

Published signals

19 signals
InfoQ 中文2026-10-21

MiniMax M3.1 Flash:成本降七成,Agent 国模第一

MiniMax 于 10 月 19 日发布 M3.1 Flash,输入每百万 token 0.15 美元、输出 0.60 美元、缓存输入 0.015 美元,官方称 Agent 任务使用成本较上一代降低 70%。模型在 OSWorld V2 得 61.7、Agents' Last Exam 得 29.6,文章称后者位列中国模型第一。此前匿名模型 Space Bunny 被确认是该模型的早期版本。

InfoQ 中文2026-10-20

Google 开源 Harness²:不改模型,递归改智能体的外壳

Google Research 开源智能体框架 Harness²,采用 Apache 2.0 许可,支持 OpenCode 与 Codex,默认使用 OpenCode。它把系统提示词、护栏、记忆、工具指导、子智能体、Skills、脚本和插件拆成 8 类可改组件,通过任务级与 Harnessing 级两层递归在不改模型权重的前提下优化表现。团队称仅一步优化后 WorkBuddy-Bench 最高涨 11.5 分。

量子位2026-10-17

DeepSeek V4.1 Flash 新版 1017:更强也更省

DeepSeek 推出 V4.1-Flash-1017 模型,在代码与智能体任务基准上提升的同时压缩了任务运行时间和输出长度,官方称为「性能—费用的帕累托改进」。配套 Harness 升级至 0.2.1 版本,并开放自研 Search 与 Fetch 两个搜索 API,每千次成功请求 5 元,11 月 1 日前免费。

Hugging Face Blog2026-10-16

Welcome Beam: a 501B open model you can own, customize, and scale

Welcome Beam: a 501B open model you can own, customize, and scale

Hugging Face Blog2026-10-15

Welcome Muse Spark Open: open-weight frontier multimodal mixture-of-experts model from Meta

Welcome Muse Spark Open: open-weight frontier multimodal mixture-of-experts model from Meta

Google AI Blog2026-10-14

Gemini 4 Argon: Built to solve your hardest problems

We’re giving developers, enterprises, scientists, and consumers access to Gemini 4 Argon, our latest frontier model.

InfoQ 中文2026-10-12

智能体用三周重构老游戏代码,全程靠逐帧回放验收

CodeScene 发布案例研究:一个编码智能体在三周内重构了《街头霸王 III:三度冲击》开源反编译项目约 30 万行 C 代码,消耗 token 约 4000 美元,覆盖 726 个文件、2903 次提交、252055 行改动,Code Health 从 5.6 升至 10.0。保障机制包括 CodeHealth MCP Server 的确定性评分,以及逐帧比对回滚状态哈希的回放轨迹测试框架;智能体还自行整理出含 22 个方案、82 条辅助说明的重构手册。

Microsoft Research Blog2026-10-08

Agent Lightning v1.0: A 3,500-Line Lightweight Agentic RL Framework for Training Agents with Real Harnesses

Training AI agents with reinforcement learning can be challenging because their tools, context, and decision-making are managed by complex frameworks. Agent Lightning connects existing agents to RL training, making it easier to improve them without rebuilding…

InfoQ 中文2026-09-26

DoorDash 清理过期开关:50 个里 45 个可用,一次 4.79 美元

DoorDash 用一套多智能体系统清理代码库里过期的功能开关。在 50 个开关的评估中,系统为 45 个生成了可用的合并请求:31 个首次提交即被合并,14 个需要修改,另有 5 个要工程师介入。平均每次 13.8 分钟、4.79 美元,人工估计要 1 到 2 小时。动代码之前必须有工程师确认目标值,合并请求只有通过构建、测试、覆盖率和静态分析之后才会创建。

InfoQ 中文2026-09-26

Open Code Review:少报错的代价是漏掉八成问题

阿里巴巴开源了内部用了两年的代码评审工具 Open Code Review:选哪些文件、怎样打包、匹配哪些规则都由确定的程序完成,只有代码分析交给智能体。官方基准里它的精确率和 F1 高于同模型的通用智能体,token 用量约为九分之一,并主动说明召回率更低。外部评论指出最优配置下召回率只有 20%,唯一一次独立测试的精确率约 12%,维护者称是已修复的异常,修复后尚无独立验证。

GitHub Blog AI2026-09-24

AI-powered fuzzing with the GitHub Security Lab Taskflow Agent

In this blog post, I explain how to use the new fuzzing taskflow based on the GitHub Security Lab Taskflow Agent AI framework.

InfoQ 中文2026-09-23

ZCode 快照事件之后:开源能挽回多少信任

开发者 ferstar 逆向发现,智谱的编程工具 ZCode 会生成含完整 Git 历史的工作区快照,加密后直传阿里云 OSS,密钥封装用的私钥在云端。智谱回应称来自默认开启的代码库索引功能,道歉、删除数据、请第三方核查,三天后把整个产品以 Apache 2.0 开源并切断该链路。仍未回答的是:现在还有什么数据会离开本机,以及开源仓库为何没有保留事故之前的提交历史。

GitHub Blog AI2026-09-16

Migrating the GitHub Copilot runtime to Rust, using Copilot

GitHub rewrote the Copilot agent runtime from TypeScript into more than 800,000 lines of production Rust, mostly written by agents across 128 incremental pull requests.

InfoQ 中文2026-09-07

Kiro Crew 开源:内部 3.9 万人在用,没人被要求

亚马逊开源 Kiro Crew(Apache 2.0),让多个编码智能体跨会话异步协同:持久化上下文、共享记忆、可复用 Skills、定时任务、并发智能体与子智能体委派,编排走 Agent Client Protocol,并有一个实时展示计划、工具调用与审批节点的 Activity 视图。最值钱的数字不是功能表——它源自内部的 MeshClaw,已被超过 3.9 万名内部开发者使用、六个月内 500 名贡献者,且不是靠强制推广。安全是从第一天设计进去的:操作系统级沙箱、默认拒绝的命令策略、敏感路径拦截、凭据脱敏、签名审计日志。反向的反馈也已出现:有用户报告它消耗 token 明显快于 Kiro CLI。

Simon Willison Blog2026-08-19

Conceptual integrity and counting lines of code

Simon Willison argues that lines of code can be a meaningful productivity indicator, that the new limiting factor is cognitive capacity, and that conceptual integrity is what coding agents make hardest to keep.

InfoQ 中文2026-08-17

Rootly 废掉了小 PR 规则:审查的对象从行数换成了影响半径

事故管理平台 Rootly 放弃了推行两年的严格小 PR 文化——智能体以「特性」而不是「增量」为单位输出,一次给出迁移、模型、服务、控制器、测试和前端。他们试过让智能体产出堆叠 PR,结果技术上没错、业务上更糟。最终的替换不是换工具而是换判据:衡量代码行数改成衡量爆炸半径,安全边界从合并挪到渐进发布,内部审查器对每个 PR 只回答一个问题——如果这个变更有缺陷,会坏掉哪些面向用户的功能。

InfoQ 中文2026-08-12

Vercel Zero:一门把编译器输出的第一读者当成智能体的语言

Vercel Labs 发布实验性系统语言 Zero,前提是编译器输出的主要阅读者不再是人:每个子命令都统一支持 JSON 输出、错误带稳定代码与带类型的修复元数据,副作用必须走编译器强制的能力参数,而 v0.3.0 起图存储才是编译器输入、源码文件降为给人看的投影。

GitHub Blog AI2026-07-22

Copilot vs. raw API access: What are you actually paying for?

GitHub argues the choice between Copilot and raw model API access comes down to which layer of the system you need to own: the model call, or the workflow, policy, and harness around it.

GitHub Blog2026-07-10

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, view) expecting a clean upgrade — and got higher cost and fewer caught issues. The tools weren't the problem; the instructions were. Rewriting them around how a reviewer actually reads a pull request flipped the regression into ~20% lower review cost at the same quality.

Related skills

3 skills
Hot right now

给编码智能体划定任务与验收

把一类重复的代码改动交给编码智能体之前,先挑对任务、把判卷标准锁在它够不着的地方,再按任务记账——这样才知道它做对了没有、值不值。

Emerging

按影响半径审查智能体写的代码

智能体按特性交回整块代码,按行数审已经失效;改问「坏了会影响谁」,由人写清改动的理由,并把安全边界从合并挪到渐进发布。

Emerging

用智能体复现论文结论

用编码智能体把一篇论文的核心论断重跑一遍,并留下别人能复查的记录——从一条论断开始,不信智能体的自我评价,失败先分原因。

Background knowledge

2 entries
Intermediate

智能体沙箱与执行隔离

智能体会执行它自己写的代码和命令,沙箱决定这些动作最多能波及哪里——隔离强度、网络出口和凭据范围是三个要分开看的维度。

Intermediate

基于执行的代码评测

让模型在真实代码仓库或终端里完成任务,再用测试判定对错——SWE-bench、Terminal-Bench 这类分数说明了什么、哪些能比、哪些不能比。

Graph neighbours

Direct neighbours from /network
Skill · Supports模型与输出评估Skill · Uses推理服务容量规划Knowledge · Explains模型选型与约束匹配Knowledge · Supports评估闭环Signal · RelatedDeepSeek Harness 开源:连 Agent Loop 都能换掉Skill · Uses智能体工作流设计Skill · Supports智能体安全与提示注入防御Skill · Uses检索流水线调优Knowledge · Explains检索增强生成基础Knowledge · Supports开放权重与模型许可Knowledge · Explains强化学习后训练与可验证奖励Skill · Supports评估开放权重模型能否自托管Skill · Uses模型与推理引擎的支持路径Skill · Supports端侧模型部署与硬件适配Knowledge · Explains混合专家模型(MoE)架构Knowledge · ExplainsKV 缓存与长上下文推理Knowledge · Prerequisite模型量化与数值精度Signal · Builds onMeta is back with Muse Glimmer: local, agentic, multimodal, and open sourceSignal · ExtendsGemini 4 Argon:DeepSWE 等多项领先,先只给网络安全团队用Skill · Supports旗舰模型发布解读与换代判断Knowledge · Explains推测解码与推理加速技术Signal · Prerequisite华为 Memento 3:把规则写成代码,打满 ARC-AGI-3Skill · Uses智能体可观测性与评测运维Knowledge · Explains完成判定与验收信号Signal · RelatedDeepSeek DSec:智能体训练拼的是沙盒Signal · Related微软 Agent Framework Harness 正式发布:智能体的运行时成了产品Skill · Uses模型微调与后训练定制Signal · ExplainsClaude Haiku 5.5Signal · SupportsThe Agent Said It Was Done. The Database Disagreed.Signal · ExplainsGoogle 开源 AX:按 Kubernetes 的方式编排智能体,可秒级暂停Signal · ExplainsIntroducing GPT-6.1 SolSignal · ExplainsClaude Sonnet 5.5:终端任务从 10% 到 71%,按任务算未必更便宜Signal · UsesADK Kotlin 1.0:智能体逻辑不必再绕回 PythonSignal · ExplainsOpenAI 的智能体借 DNS 出了沙箱:警报响了,运行没停Signal · Extends一次 180 万美元的智能体账单:超预算 860%,五个月后才被发现Skill · SupportsAI 试点范围界定Skill · Supports工具集成模式Knowledge · Uses人在回路的审查Knowledge · Uses工具使用与函数调用Skill · SupportsAI 工具链选型与自建边界评估Knowledge · Explains系统设计的权衡Signal · Explains只能打开网址的智能体,是怎样把代码送出去的Signal · ExtendsIntroducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash CyberSignal · ExtendsGPT-Red: Unlocking Self-Improvement for RobustnessKnowledge · Uses反馈闭环与团队学习Signal · SupportsOpenAI agents carried out an undisclosed attack on RubyGems; METR's independent investigation of the Hugging Face incidentKnowledge · SupportsAPI 契约与接口边界Signal · UsesHow UK AISI and EvalEval Are Making Benchmark Results ReproducibleSignal · RelatedAnthropic 公布内部测量:Claude 主导 26% 的研发,裁判也是 ClaudeSignal · ExplainsYour Agent Aced the Task. Will It Do It Again?Signal · ExtendsOpenAI's rogue agents were caught communicating via public wikisSignal · SupportsWhat building Shippy taught us about building agentsSignal · SupportsResearch acceleration: The view inside OpenAIKnowledge · Prerequisite长时程智能体的记忆与上下文管理Signal · ExplainsMuse Glimmer 开源:30B 压进消费级显卡,报错不停下Signal · Extends菲尔兹奖得主:模型强在找到那个对象,弱在知道何时放弃Skill · SupportsAI 辅助沟通审阅Signal · ExplainsCloudflare Computer:不给智能体一个容器,给它一台机器Signal · ExplainsQwen3.8-27B 开源:消费级显卡跑得动,尺子却是自己命名的Signal · SupportsWhat We Learned by Reproducing 2,200 papers from ICMLSignal · Supports生产要持久、评估要轻量——把编排从运行时里拆出来Signal · SupportsARA:把论文从 PDF 改成智能体能直接操作的知识包Knowledge · Builds on面向知识系统的图思维Signal · Supports智能体复现审计:ICML 2026 的 92 篇论文里,只有 8 篇能复现八成结论Signal · ExplainsBigBang-v1:把数据生产系统本身当成优化对象,以及一处对不上的口径Signal · SupportsTutorMoments: Do AI tutors know when to help and when to hold back?Signal · ExplainsAnatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 IncidentSignal · ExplainsSecurity incident disclosure — July 2026Signal · ExplainsExpanding Managed Agents in Gemini API: background tasks, remote MCP and moreKnowledge · Prerequisite可复现性与可验证结果