Skip to content

#Tutorials/practice

0 items today
10/5Mon
  2. DEV Community · MCP30

    如何用 Omni Gateway 策略治理 AI 智能体流量：MCP、A2A 与 LLM 三层防护

    Omni Gateway 位于每个 MCP server、智能体间调用和 LLM 请求之前，用策略集分别治理三类流量。MCP 侧通过 MCP Global Access 只放行 Part 4 的五个工具，并用 MCP Attribute-Based Access Control 按调用方身份收窄权限，Intake 智能体看不到 import_transport_to_qa。

    Awaiting translation

  4. DEV Community · Codex78

    How I Stopped Codex from Burning Through My Usage Quota

    The author found that two Codex browser automation tasks consumed 170,123 and 110,180 tokens respectively, so they set out to control usage through model selection, configuration files, and task splitting.

    Why it matters: Drawing on real measurements where two browser tasks burned through hundreds of thousands of tokens, the author shares a quota-saving approach: switch models and configurations based on task difficulty.

  9. DEV Community · Vibe Coding22

    What goes wrong with apps launched on Vibe Coding: continuevibe takes over maintenance

    The small Korean team continuevibe offers maintenance services for products launched on Vibe Coding, with a process of diagnosing the problem, scoping the fix, fixing the code, and testing. From their interviews, they found that the most common post-launch issues are severe bugs that drive users away, such as broken login and data loss, along with fixing one thing and breaking another—failures that non-developers find hard to describe.

  14. DEV Community · Vibe Coding32

    Taking my vibe-coded travel app Le Voyage to New Orleans: when the app goes on vacation too

    The author took his travel-planning app Le Voyage—which he built with the AI app builder vibecode over nearly a year, with more than a hundred GitHub commits—to New Orleans. The two of them saved 68 restaurants, bars, cafés, and attractions in the app. In the week before the trip, he added features like remembering filter and map preferences per itinerary, sorting, and a confirmation step before deletion, and realized that what was really missing wasn't AI recommendations but the ability to pick three options from an existing list based on location and needs.

10/4Sun
  2. DEV Community · Claude Code85

    How to Stop an AI Coding Agent from Declaring a Task Done Too Early

    The author runs a fully autonomous implementation system where an orchestrator hands out tasks to parallel implementation agents (built on Claude Code). At first, agents could mark their own tasks as complete, which led to problems like tests never being run, acceptance criteria not being met, assertions loosened to make tests pass, and hardcoded return values.

    Why it matters: The author solved the problem of agents declaring completion too early with a three-layer design: checkable acceptance criteria, completion reports backed by evidence, and read-only validation agents.

  3. DEV Community · AI Coding26

    2026 年 7 款最佳 GitHub Copilot 替代品实测：免费额度与价格对比

    2026 年 10 月的实测对比显示，Cursor 以 4.4 分成为多数开发者的最佳 GitHub Copilot 替代品，Hobby 免费、Pro 每月 20 美元；Zed AI 和 Continue 在免费额度上最强，Claude Code 智能体得分最高（5.0）但无免费层。GitHub Copilot 仍以每月 10 美元的 Pro 席位守住性价比基线，其免费层仅支持自动模型选择。

    Awaiting translation

  4. Habr · Cursor22

    Cursor 如何取代我的 Stack Overflow、加速调试并让我重拾业余项目

    一位开发者在公司要求学习 Cursor 后，用它几分钟内生成 Django 网店模板、视图和 URL，并完成爬虫、入库与分析等原本耗时数小时甚至数天的任务。他把 Cursor 用于 OfferIntel、股票 RAG 和 Author.Today 阅读统计三个业余项目，认为 AI 没有取代他，而是消除重复劳动，部分场景开发速度提升数十倍。

    Awaiting translation

  5. Habr · Вайбкодинг34

    Vibe coding an ERP with AI—can it replace buying off-the-shelf software?

    With 15 years of experience building the СКИФ warehouse system, the author argues AI can write a working core for inventory and financial accounting, but business rules like negative stock, backdated documents, and НДС rounding still need a human to decide, and AI often gets Russian-specific integration details wrong—НДС, УПД, labeling, ККМ, ЭДО. A minimal viable core is roughly a few thousand lines of code, 15–25 tables, and dozens of screens, and it only really matures in the first six months after launch. AI doesn't take responsibility for errors, while a vendor will fix bugs for free.

  10. 效率火箭36

    Vivaldi 8.2 把 Chrome 扩展带到 Android

    Vivaldi 8.2 已把 Chrome 扩展支持带进 Android 正式版，手机和平板可直接通过 Chrome Web Store 安装扩展。安装时需开启「桌面版网站」选项，扩展不会随 Google 账号自动同步，每台设备需单独安装，部分数据可经 Vivaldi Sync 同步。该支持目前仅限 Android，iOS 因平台扩展机制限制未提供。

    Awaiting translation

  11. 宝玉82

    Drawing on a podcast episode, Baoyu walks through how Lauren Tan, who works on Grok Bot at SpaceXAI, merged 2500 PRs in a single month: at night she lets the AI check and merge on its own, then spot-checks the next morning instead of reviewing each one.

    Quotedlauren@poteto

    i had a lot of fun chatting with @mattpocockuk today about how i was able to land 2,500 PRs last month! Matt is a wonderful interviewer so i think the interview turned out really interesting both of our skill plugins work great together, so i recommend giving both a try and picking the best skills that suit your workflow https://www.youtube.com/watch?v=MN9dGgmLyso

    Why it matters: Using Lauren Tan's practice of merging 2500 PRs in a month, Baoyu explains that skipping individual reviews rests on validation Skills and rule constraints, and lays out the conditions under which he'd apply the same approach.

  15. 佬刘AI78

    Planning with GPT, Execution with DeepSeek: A Cost-Saving Two-Model Workflow

    The author has GPT (gpt-6.1-sol, reasoning tier high) handle planning, key decisions, and acceptance in Codex, and calls DeepSeek-V4.1-Flash through the official DeepSeek Harness to write code, run experiments, and fix bugs—together producing a local PDF toolkit with five working features.

    Why it matters: The author splits the work between GPT for planning and DeepSeek for execution to get a PDF toolkit running, and shares three prompts plus cache usage data that can carry over to cutting costs on long tasks.

  17. 苏三说技术34

    微信为什么使用SQLite保存聊天记录？

    微信聊天记录默认不上云，而是以SQLite文件形式存在用户手机本地，服务器只负责转发消息。腾讯在SQLite之上封装了WCDB框架，加入SQLCipher的AES-256-CBC加密，并借助FTS5倒排索引实现快速搜索，实测1万条长文本查询中位响应约0.14毫秒，比LIKE快15倍。数据库文件超过约240MB时会自动拆分。

    Awaiting translation

  20. Kondasamy Jayaraman · Engineering Blog20

    TIL: byte-by-byte file comparison in shell scripts with cmp

    To check whether two files are identical in a shell script, use cmp instead of diff: cmp compares byte by byte and stops at the first mismatch, finishing in milliseconds even for GB-sized files, while diff has to parse the whole file line by line. cmp's -s silent mode reports the result through its exit code: 0 means the files are identical, 1 means they differ, and 2 means a file is missing or unreadable. cmp suits binary files and automation scripts, and ships by default in POSIX environments.

  22. Ben Holmes58

    Ben Holmes 公开了驱动其软件工厂的多智能体系统：需求来自 Linear issue 或 Slack 讨论，分诊智能体先调研并决定是直接实现还是提问，实现智能体负责构建、必要时先与子智能体协作产出 spec，验证子智能体对实现结果做端到端测试，代码评审智能体与实现智能体循环几轮后再交人工评审上线，监控自动化则响应告警并创建 issue。

    Awaiting translation

  23. Hacker News · AI 编程经验问答38

    AI 编程是否让软件库建设变得过时？

    有开发者提出，随着最新一轮模型发布，AI 已能在生成应用时把几乎任何库的功能内联生成，因此软件库建设可能不再必要。过去把功能拆成定义清晰的模块、增量发布并复用代码，是在构建耗时较长时的必要做法，如今这些做法的重要性明显下降。该问题目前只是提问与讨论，尚无结论。

    Awaiting translation