<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>LLM on Farid</title><link>https://polymatx.dev/blog/tags/llm/</link><description>Recent content in LLM on Farid</description><generator>Hugo</generator><language>en</language><lastBuildDate>Sun, 05 Jul 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://polymatx.dev/blog/tags/llm/index.xml" rel="self" type="application/rss+xml"/><item><title>Weave: I Built a TypeScript-Native Agent Orchestrator (And Why You Might Want One)</title><link>https://polymatx.dev/blog/posts/weave-typescript-agent-orchestrator/</link><pubDate>Sun, 05 Jul 2026 00:00:00 +0000</pubDate><guid>https://polymatx.dev/blog/posts/weave-typescript-agent-orchestrator/</guid><description>&lt;p&gt;After months of building LLM agent pipelines for real projects, I kept hitting the same three walls: my graph state degraded into &lt;code&gt;any&lt;/code&gt; soup, wiring MCP servers into agents meant writing the same glue code over and over, and answering &amp;ldquo;why did this run cost $4?&amp;rdquo; required either a SaaS observability subscription or grepping JSON logs at 3 a.m.&lt;/p&gt;
&lt;p&gt;So I built &lt;a href="https://github.com/polymatx/weave" class="external-link" target="_blank" rel="noopener"&gt;weave&lt;/a&gt; — a small, opinionated framework for multi-agent workflows in TypeScript. It&amp;rsquo;s MIT licensed, published on npm, and the core is about 600 lines of code. This post is a tour of what it does and why I made the choices I made.&lt;/p&gt;</description></item></channel></rss>