<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Posts on Shui Song Luar</title><link>https://luarss.dev/posts/</link><description>Recent content in Posts on Shui Song Luar</description><image><title>Shui Song Luar</title><url>https://luarss.dev/img/og-default.jpg</url><link>https://luarss.dev/img/og-default.jpg</link></image><generator>Hugo -- 0.161.1</generator><language>en-us</language><copyright>2020–2026 Shui Song Luar · Content licensed under CC BY-NC 4.0 · llms.txt</copyright><lastBuildDate>Thu, 12 Nov 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://luarss.dev/posts/index.xml" rel="self" type="application/rss+xml"/><item><title>Building the CyberSG Atlas Chatbot in Four Weeks</title><link>https://luarss.dev/posts/cybersg-atlas-chatbot/</link><pubDate>Thu, 12 Nov 2026 00:00:00 +0000</pubDate><guid>https://luarss.dev/posts/cybersg-atlas-chatbot/</guid><description>How we built and red-teamed a catalogue-grounded RAG chatbot for CyberSG Atlas, a directory of 200+ Singapore cybersecurity companies, what we moved from the prompt into code, and what the first traces taught us.</description></item><item><title>Supply Chain Hygiene for Claude Code Plugins</title><link>https://luarss.dev/posts/supply-chain-hygiene-claude-plugins/</link><pubDate>Mon, 19 Oct 2026 00:00:00 +0000</pubDate><guid>https://luarss.dev/posts/supply-chain-hygiene-claude-plugins/</guid><description>How to pin Claude Code plugins, GitHub Actions, and zsh plugins to verified commit SHAs, and automate update checks before anything lands in your dev environment.</description></item><item><title>OpenROAD MCP: Let AI Close Your Timing</title><link>https://luarss.dev/posts/openroad-mcp-let-ai-close-your-timing/</link><pubDate>Thu, 19 Feb 2026 00:00:00 +0000</pubDate><guid>https://luarss.dev/posts/openroad-mcp-let-ai-close-your-timing/</guid><description>OpenROAD MCP is a server that lets AI assistants like Claude Code talk directly to a live OpenROAD process, so timing analysis and constraint fixes happen in one conversation.</description></item><item><title>Building Robust AI Research Agents: A tale of five specialists</title><link>https://luarss.dev/posts/building-robust-ai-research-agents/</link><pubDate>Thu, 20 Nov 2025 00:00:00 +0000</pubDate><guid>https://luarss.dev/posts/building-robust-ai-research-agents/</guid><description>Four architectural iterations of building a deep research workflow for VCs, from a single agent to five specialized agents with hierarchical citations, MCP integration, and iterative refinement.</description></item><item><title>Multi-Agent Systems Are 23x Cheaper Than Deep Research Models with Higher Quality</title><link>https://luarss.dev/posts/multi-agent-systems-23x-cheaper/</link><pubDate>Thu, 13 Nov 2025 00:00:00 +0000</pubDate><guid>https://luarss.dev/posts/multi-agent-systems-23x-cheaper/</guid><description>A benchmark of an 11-agent hypothesis generation system against GPT-4o-mini with tools and o4-mini-deep-research across 5 scientific domains. The multi-agent system won or tied in all 5 at a fraction of the cost.</description></item></channel></rss>