一篇基于本地知识库证据自动综合的博客草稿。
大多数系统停留在“收集与检索”,缺乏“编译与回灌”。真正的升级,不是多存几篇文章,而是把每次提问都变成可复用资产。
raw/2026-03-07-blog-第-3473-期-这就是如何利用大型语言模型来帮助我编写代码的方法-c33013.md(score=61)raw/2026-03-07-blog-django-创造者-simon-willison-分享-我如何使用-llm-帮我写代码-5365b2.md(score=60)raw/2026-03-07-blog-my-llm-coding-workflow-going-into-2026-015408.md(score=51)raw/2026-03-27-2026-03-27-autoresearchclaw-全自主研究管线.md(score=45)raw/2026-03-07-blog-how-to-turn-claude-code-into-a-domain-specific-coding-agent-9e2421.md(score=41)raw/2026-03-07-blog-conversation-llms-and-the-what-how-loop-9b04fc.md(score=39)raw/2026-03-07-blog-how-to-not-be-overwhelmed-by-ai-a-developer-s-guide-to-using-e8aaeb.md(score=36)raw/2026-03-07-blog-how-to-write-a-good-spec-for-ai-agents-000c9c.md(score=35)知识库真正的价值,不在于你存了多少,而在于系统是否会随着每次提问变得更强。
[blog-import]: 8migrated: 4research: 1llm: 1knowledge-base: 1agent: 1raw/2026-03-07-blog-第-3473-期-这就是如何利用大型语言模型来帮助我编写代码的方法-c33013.md · score=61 · --- title: "【第 3473 期】这就是如何利用大型语言模型来帮助我编写代码的方法" url: --- # 【第 3473 期】这就是如何利用大型语言模型来帮助我编写代码的方法 前言 分享如何利用大型语言模型(LLMs)来辅助编写代码,强调了这一过程的困难性和非直观性,并提供了一系列实用的技巧和经验。今日前端早读课文章由 @Simon Willison 分享,@飘飘翻译。 译文从这开始~~ 在讨论如何使用大型语言模型(LLM)编写代码时,总会有开发者表示他们的体验并不理想。他们常常疑惑自己是不是做错了什么 ------ 为什么有些人能取得不错的成果,而自己却收效甚微? 使用 LLM 编写代码既困难又不直观。要真正掌握它的优...raw/2026-03-07-blog-django-创造者-simon-willison-分享-我如何使用-llm-帮我写代码-5365b2.md · score=60 · --- title: "Django 创造者 Simon Willison 分享:我如何使用 LLM 帮我写代码" url: --- # Django 创造者 Simon Willison 分享:我如何使用 LLM 帮我写代码 近段时间,著名 AI 科学家 Andrej Karpathy 提出的氛围编程(vibe coding)是 AI 领域的一大热门话题。简单来说,氛围编程就是鼓励开发者忘掉代码,进入开发的氛围之中。更简单地讲,就是向 LLM 提出需求,然后「全部接受」即可。 当然,这种使用 LLM 编程的方法过于简单粗暴,或许并不适合用来开发一些更加复杂和精细的项目。近日,探索和发表数据的开源工具 Datasette 的...raw/2026-03-07-blog-my-llm-coding-workflow-going-into-2026-015408.md · score=51 · --- title: "My LLM coding workflow going into 2026" url: --- # My LLM coding workflow going into 2026 AI coding assistants became game-changers this year, but harnessing them effectively takes skill and structure. These tools dramatically increased what LLMs can do for real-world coding, and many developers (myself...raw/2026-03-27-2026-03-27-autoresearchclaw-全自主研究管线.md · score=45 · # AutoResearchClaw:从想法到论文的全自主研究管线 > 项目:aiming-lab/AutoResearchClaw > 来源: > Stars:8,896 · Forks:945 · Language:Python · License:MIT > 更新:2026-03-26 > 标签:autonomous-research, multi-agent-debate, self-evolving, openclaw --- ## Chat an Idea. Get a Paper. Fully Autonomous & Self-Evolving. --- [](docs/showcase/SHOWCASE.md) *...raw/2026-03-07-blog-how-to-turn-claude-code-into-a-domain-specific-coding-agent-9e2421.md · score=41 · --- title: "How to turn Claude Code into a domain specific coding agent" url: --- # How to turn Claude Code into a domain specific coding agent Authored by: Coding agents are great at writing code that uses popular libraries on which LLMs have been heavily trained on. But point them to a custom library, a new version o...raw/2026-03-07-blog-conversation-llms-and-the-what-how-loop-9b04fc.md · score=39 · --- title: "Conversation: LLMs and the what/how loop" url: --- # Conversation: LLMs and the what/how loop Unmesh The primary challenge of software development is to build systems that survive change. People new to programming, and people who don't code for a living but hire those who do, often think of programming as a...raw/2026-03-07-blog-how-to-not-be-overwhelmed-by-ai-a-developer-s-guide-to-using-e8aaeb.md · score=36 · --- title: "How to Not Be Overwhelmed by AI – A Developer’s Guide to Using AI Tools Effectively" url: --- # How to Not Be Overwhelmed by AI – A Developer’s Guide to Using AI Tools Effectively If you're a developer, you'll likely want to use AI to boost your productivity and help you save time on menial, repetitive task...raw/2026-03-07-blog-how-to-write-a-good-spec-for-ai-agents-000c9c.md · score=35 · --- title: "How to write a good spec for AI agents" url: --- # How to write a good spec for AI agents **TL;DR: Aim for a clear spec covering just enough nuance (this may include structure, style, testing, boundaries) to guide the AI without overwhelming it. Break large tasks into smaller ones vs. keeping everything in ...raw/2026-04-03-2026-04-03-karpathy-llm-knowledge-bases-个人知识库构建方法.md · score=33 · # LLM Knowledge Bases — Karpathy 的个人知识库构建方法 > 作者:Andrej Karpathy > 类型:方法论 / 工作流 > 日期:2026-04-03 --- ## 核心理念 用 LLM 构建个人知识库。大部分 token 消耗不是在操作代码,而是在操作知识(以 Markdown 和图片形式存储)。 ## 完整工作流 ### 1. Data Ingest(数据摄入) - 将源文档(文章、论文、仓库、数据集、图片等)索引到 raw/ 目录 - 用 LLM 增量"编译"成一个 wiki——就是一个目录结构里的 .md 文件集合 - Wiki 包含:所有 raw/ 数据的摘要、反向链...raw/2026-03-07-blog-context-engineering-bf3dc3.md · score=30 · --- title: "Context Engineering" url: --- # Context Engineering ### TL;DR {#tldr} Agents need context to perform tasks. Context engineering is the art and science of filling the context window with just the right information at each step of an agent's trajectory. In this post, we break down some common strategies --- *...raw/2026-03-28-如何从零开始实现一个-ai-agent-框架-理论-实践.md · score=27 · # 如何从零开始实现一个 AI Agent 框架(理论+实践) > 公众号: 腾讯云开发者 > 发布时间: 2026-02-26 08:46 > 原文链接: --- 关注腾讯云开发者,一手技术干货提前解锁👇 > 人类最重要的 affordance 是“手”,而 AI 最重要的 affordance 可能是“代码”。 > > Shunyu Yao 年初火爆现在热度依旧的OpenClaw为AI Agent带来了新的想象。如果说2025是AI Agent智能体元年,想必2026将会是AI Agent真正商用化的开端,而AI Agent商用化的前提是各行各业开始落地实际商业领域的AI Agent。 作为一名程序员/工程师,思考的更多的可能是...raw/2026-04-03-2026-04-03-karpathy-llm-knowledge-bases-个人知识库构建方法-1.md · score=27 · # LLM Knowledge Bases — Karpathy 的个人知识库构建方法 > 作者:Andrej Karpathy > 类型:方法论 / 工作流 > 日期:2026-04-03 --- ## 核心理念 用 LLM 构建个人知识库。大部分 token 消耗不是在操作代码,而是在操作知识(以 Markdown 和图片形式存储)。 ## 完整工作流 ### 1. Data Ingest(数据摄入) - 将源文档(文章、论文、仓库、数据集、图片等)索引到 raw/ 目录 - 用 LLM 增量"编译"成一个 wiki——就是一个目录结构里的 .md 文件集合 - Wiki 包含:所有 raw/ 数据的摘要、反向链...