<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>首页 on 时光的枳树</title><link>https://www.timlant.cn/</link><description>Recent content in 首页 on 时光的枳树</description><generator>Hugo</generator><language>zh-cn</language><lastBuildDate>Tue, 21 Apr 2026 20:40:00 +0800</lastBuildDate><atom:link href="https://www.timlant.cn/index.xml" rel="self" type="application/rss+xml"/><item><title>Go 后端实战：从 Handler 到 Repository 的清晰分层</title><link>https://www.timlant.cn/posts/go-clean-architecture-practice/</link><pubDate>Tue, 21 Apr 2026 20:40:00 +0800</pubDate><guid>https://www.timlant.cn/posts/go-clean-architecture-practice/</guid><description>这篇文章用一个订单服务示例，讲清楚 Go 后端常见的分层方式，以及如何避免业务逻辑散落。</description></item><item><title>Go 并发模型在后端里的正确打开方式：Worker Pool 与限流</title><link>https://www.timlant.cn/posts/go-concurrency-worker-pool/</link><pubDate>Mon, 20 Apr 2026 19:20:00 +0800</pubDate><guid>https://www.timlant.cn/posts/go-concurrency-worker-pool/</guid><description>并发不是越多越快。这篇文章讨论 Go 后端中 Worker Pool、超时控制和限流的组合策略。</description></item><item><title>C# 后端架构实践：在 .NET 里把业务复杂度收住</title><link>https://www.timlant.cn/posts/csharp-dotnet-clean-architecture/</link><pubDate>Sun, 19 Apr 2026 21:00:00 +0800</pubDate><guid>https://www.timlant.cn/posts/csharp-dotnet-clean-architecture/</guid><description>围绕 ASP.NET Core 项目，讨论 Application、Domain、Infrastructure 分层的实战取舍。</description></item><item><title>C# 后端性能排查：从慢接口到瓶颈定位的一条路径</title><link>https://www.timlant.cn/posts/csharp-performance-diagnostics/</link><pubDate>Sat, 18 Apr 2026 22:10:00 +0800</pubDate><guid>https://www.timlant.cn/posts/csharp-performance-diagnostics/</guid><description>性能问题不可怕，可怕的是没有路径。本文给出 .NET 接口变慢时的排查顺序和观察指标。</description></item><item><title>个人技术站点</title><link>https://www.timlant.cn/projects/portfolio-site/</link><pubDate>Sat, 18 Apr 2026 11:00:00 +0800</pubDate><guid>https://www.timlant.cn/projects/portfolio-site/</guid><description>&lt;p&gt;这是一个面向开发者与技术写作者的站点模板。&lt;/p&gt;
&lt;p&gt;目标不是功能堆叠，而是通过稳定的内容结构持续积累长期价值。&lt;/p&gt;</description></item><item><title>Transformer 原理入门：用直觉理解 Attention 在做什么</title><link>https://www.timlant.cn/posts/transformer-attention-intuition/</link><pubDate>Fri, 17 Apr 2026 20:00:00 +0800</pubDate><guid>https://www.timlant.cn/posts/transformer-attention-intuition/</guid><description>这篇文章不堆公式，先用直觉讲清楚 Query、Key、Value 和多头注意力的意义。</description></item><item><title>Transformer 训练与推理：为什么训练快、生成却慢</title><link>https://www.timlant.cn/posts/transformer-training-and-inference/</link><pubDate>Thu, 16 Apr 2026 21:30:00 +0800</pubDate><guid>https://www.timlant.cn/posts/transformer-training-and-inference/</guid><description>同样是 Transformer，训练和推理的计算模式差异很大。本文解释其背后的原因与工程影响。</description></item><item><title>前端渲染策略怎么选：CSR、SSR、SSG 与混合架构实战</title><link>https://www.timlant.cn/posts/frontend-rendering-strategy/</link><pubDate>Wed, 15 Apr 2026 20:10:00 +0800</pubDate><guid>https://www.timlant.cn/posts/frontend-rendering-strategy/</guid><description>渲染方案没有银弹。本文从性能、SEO、团队协作和维护成本四个维度，给出可落地的选择框架。</description></item><item><title>知识笔记系统</title><link>https://www.timlant.cn/projects/notes-system/</link><pubDate>Wed, 15 Apr 2026 14:00:00 +0800</pubDate><guid>https://www.timlant.cn/projects/notes-system/</guid><description>&lt;p&gt;这类项目的价值不在技术炫技，而在于它是否真正降低了记录、整理与发布的成本。&lt;/p&gt;</description></item><item><title>前端性能实战：用 Core Web Vitals 做持续优化</title><link>https://www.timlant.cn/posts/frontend-core-web-vitals-practice/</link><pubDate>Tue, 14 Apr 2026 20:30:00 +0800</pubDate><guid>https://www.timlant.cn/posts/frontend-core-web-vitals-practice/</guid><description>性能优化不是一次性冲刺，而是可持续工程。本文给出围绕 LCP、INP、CLS 的治理路径。</description></item><item><title>前端状态管理的边界：什么时候该上全局状态，什么时候不该</title><link>https://www.timlant.cn/posts/frontend-state-management-boundary/</link><pubDate>Mon, 13 Apr 2026 21:00:00 +0800</pubDate><guid>https://www.timlant.cn/posts/frontend-state-management-boundary/</guid><description>状态管理最大的坑不是选错库，而是边界定义不清。本文讨论局部状态、服务端状态和全局状态的分工。</description></item><item><title>设计系统落地：前端团队如何把组件库用成生产力</title><link>https://www.timlant.cn/posts/frontend-design-system-governance/</link><pubDate>Sun, 12 Apr 2026 20:20:00 +0800</pubDate><guid>https://www.timlant.cn/posts/frontend-design-system-governance/</guid><description>设计系统不是做一套 UI 组件就结束，而是长期治理问题。本文从 token、组件约束和发布流程谈落地方法。</description></item><item><title>后端接口幂等与重试：把重复请求变成可控行为</title><link>https://www.timlant.cn/posts/backend-idempotency-and-retry/</link><pubDate>Sat, 11 Apr 2026 21:00:00 +0800</pubDate><guid>https://www.timlant.cn/posts/backend-idempotency-and-retry/</guid><description>支付、下单、回调场景里，重复请求很常见。本文讨论幂等键设计、重试策略与一致性边界。</description></item><item><title>后端可观测性从 0 到 1：日志、指标、链路如何协同</title><link>https://www.timlant.cn/posts/backend-observability-from-zero/</link><pubDate>Fri, 10 Apr 2026 20:40:00 +0800</pubDate><guid>https://www.timlant.cn/posts/backend-observability-from-zero/</guid><description>可观测性不是装几个监控就结束。本文给出后端团队可直接落地的日志、指标与链路协作方案。</description></item><item><title>AI 应用怎么选：RAG、微调还是两者结合</title><link>https://www.timlant.cn/posts/ai-rag-vs-finetune/</link><pubDate>Thu, 09 Apr 2026 20:50:00 +0800</pubDate><guid>https://www.timlant.cn/posts/ai-rag-vs-finetune/</guid><description>很多团队在 RAG 和微调之间反复摇摆。本文给出一个按场景决策的简单框架。</description></item><item><title>LLM 应用评测体系：没有评测，就没有稳定迭代</title><link>https://www.timlant.cn/posts/ai-llm-evaluation-system/</link><pubDate>Wed, 08 Apr 2026 21:10:00 +0800</pubDate><guid>https://www.timlant.cn/posts/ai-llm-evaluation-system/</guid><description>模型效果的讨论必须可量化。本文介绍一套适合业务团队的 LLM 评测与回归机制。</description></item><item><title>联邦学习架构实战：从中心训练到跨端协同的工程路线</title><link>https://www.timlant.cn/posts/federated-learning-architecture-practice/</link><pubDate>Tue, 07 Apr 2026 20:30:00 +0800</pubDate><guid>https://www.timlant.cn/posts/federated-learning-architecture-practice/</guid><description>联邦学习不只是算法问题，更是系统工程问题。本文讨论参与方管理、训练编排、聚合策略和上线治理。</description></item><item><title>联邦学习里的 Non-IID 难题：如何让模型更稳地收敛</title><link>https://www.timlant.cn/posts/federated-learning-noniid-optimization/</link><pubDate>Mon, 06 Apr 2026 21:00:00 +0800</pubDate><guid>https://www.timlant.cn/posts/federated-learning-noniid-optimization/</guid><description>数据异质性是联邦学习的核心难点。本文从算法和工程两侧讨论缓解 Non-IID 的实用策略。</description></item><item><title>多方安全计算（MPC）入门：协议怎么选，边界怎么看</title><link>https://www.timlant.cn/posts/mpc-basics-and-protocol-selection/</link><pubDate>Sun, 05 Apr 2026 20:20:00 +0800</pubDate><guid>https://www.timlant.cn/posts/mpc-basics-and-protocol-selection/</guid><description>MPC 常被当成黑盒能力。本文用工程视角说明 MPC 的价值、代价以及协议选择方法。</description></item><item><title>MPC 工程化落地：性能、可观测性与生产运维要点</title><link>https://www.timlant.cn/posts/mpc-engineering-and-performance/</link><pubDate>Sat, 04 Apr 2026 20:45:00 +0800</pubDate><guid>https://www.timlant.cn/posts/mpc-engineering-and-performance/</guid><description>隐私计算项目从 PoC 到生产，难点在工程细节。本文总结 MPC 上线中最容易被忽视的环节。</description></item></channel></rss>