<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Vamshi Pokala on Medium]]></title>
        <description><![CDATA[Stories by Vamshi Pokala on Medium]]></description>
        <link>https://medium.com/@vamshipokala?source=rss-153e72d3ace9------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*5lS3uD4Yvtc37FHnJpjYcA.jpeg</url>
            <title>Stories by Vamshi Pokala on Medium</title>
            <link>https://medium.com/@vamshipokala?source=rss-153e72d3ace9------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 10:19:19 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@vamshipokala/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Second Brain Lite: a local AI wiki you can run in five minutes]]></title>
            <link>https://vamshipokala.medium.com/second-brain-lite-a-local-ai-wiki-you-can-run-in-five-minutes-52f7cbdd051f?source=rss-153e72d3ace9------2</link>
            <guid isPermaLink="false">https://medium.com/p/52f7cbdd051f</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[rust]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[open-source]]></category>
            <dc:creator><![CDATA[Vamshi Pokala]]></dc:creator>
            <pubDate>Mon, 11 May 2026 03:19:44 GMT</pubDate>
            <atom:updated>2026-05-11T03:28:53.019Z</atom:updated>
            <content:encoded><![CDATA[<p><em>And why I built it instead of using one of the twelve tools that already exist</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ed7wzpM_4I-z08PDOWcPrQ.png" /></figure><p>I’ve spent the last few years building AI systems for a living. RAG pipelines, agentic workflows, knowledge graphs wired up to LLMs. And for my own notes, I was using a mix of cloud tools I’d never fully trust with anything that actually mattered.</p><p>That bothered me more than it should have.</p><p>The thing that finally pushed me to build something was Andrej Karpathy’s idea of an LLM wiki — a personal knowledge base that gets smarter the more you use it, builds memory over time, and helps you actually find and connect what you know rather than just storing it. The concept is simple and honestly a bit obvious once you hear it: your notes should behave like a brain that grows, not a filing cabinet that fills up. I wanted that, but locally, with no account required and no one indexing my content in the background.</p><p>Most AI-powered note apps need your content to live on their servers. Some are upfront about that. Others aren’t. Either way, if you’re working on anything sensitive — client context, early ideas, half-formed thoughts you haven’t pressure-tested yet — you’re making a tradeoff you probably didn’t think through when you signed up.</p><p>So I built Second Brain Lite. Not to replace Notion or Obsidian. Just a local-first wiki with AI chat that runs entirely on your machine. Your documents stay on your disk. It ingests what you feed it, builds context over time, and helps you think through things without any of it leaving your machine — especially if you run it with a local LLM like Ollama.</p><p>Here’s how to get it running.</p><p><strong>What you actually need</strong></p><p>Node.js. That’s it for the browser version.</p><p>If you want the full native desktop app — which runs as a proper application rather than a browser tab — you’ll also need Rust. I’ll cover that separately. Most people can skip it for now.</p><h3>Cloning the repo</h3><pre>git clone https://github.com/vampokala/Second-Brain-App-Lite.git<br>cd Second-Brain-App-Lite</pre><h3>The part I’m actually proud of</h3><p>I got tired of projects that need four pages of setup before anything happens. So there are two scripts — one for each platform. That’s the whole interface.</p><p>On Mac or Linux:</p><pre>chmod +x run-wiki.sh<br>./run-wiki.sh</pre><p>On Windows:</p><pre>run-wiki.bat</pre><p>The script handles everything from there. It checks for Node, installs packages if node_modules is missing, then tries to launch the Tauri desktop app if Rust is installed. If it can&#39;t — no Rust, a firewall blocking localhost, a VPN doing something weird — it falls back automatically and opens the app in your browser at <a href="http://localhost:1420.">http://localhost:1420.</a></p><p>I work on a corporate laptop sometimes. The automatic fallback has saved me from a lot of unnecessary debugging.</p><h3>If you want the native desktop app</h3><p>The desktop version runs outside the browser, which is nice for focus and avoids the port conflict problem. To get it, install Rust first:</p><pre>curl --proto &#39;=https&#39; --tlsv1.2 -sSf https://sh.rustup.rs | sh</pre><p>Restart your terminal after that, then run the script again. It’ll detect Rust and launch the desktop app directly without you having to change anything.</p><h3>Your API key</h3><p>Once it’s running, go to Settings and drop in your OpenAI or Anthropic key. It’s stored encrypted on disk and goes nowhere except directly to the API when you make a request.</p><p>That’s not zero trust or anything fancy. The app is local, so the threat model is just “don’t store my key as plain text” — which is already better than what most tutorials tell you to do. If you’re querying GPT-4, those queries still leave your machine. That’s fine, but worth being clear about rather than pretending local-first means fully private.</p><h3>Adding your notes</h3><p>The sidebar handles navigation. Create a page, start writing — it’s markdown, so links between pages work the way you’d expect. The AI chat sits alongside your content and doesn’t care what you ask it. Summarize, brainstorm, poke at an argument. Because there’s no vendor deciding what the AI feature should do, it’s just a text box wired to whatever model you’ve configured.</p><h3>What I’d do differently</h3><p>I built this as an open source project partly to have something concrete to talk about when people ask me why every note tool needs an account. The answer I kept giving was “I don’t know, but you can build one that doesn’t.” So I did.</p><p>If I started over, the import story would come first. Getting notes out of Notion or Bear and into something new is where people give up before they’ve seen whether the tool is worth it. That friction is real and I haven’t solved it yet.</p><p>I’d also have been more direct earlier about the privacy model. Local-first is a meaningful thing — your data isn’t indexed by some company’s training pipeline. But it doesn’t mean the app is airgapped. If you’re using a cloud LLM, you’re trusting that provider. Worth saying plainly.</p><p>The full code is at <a href="https://github.com/vampokala/Second-Brain-App-Lite">github.com/vampokala/Second-Brain-App-Lite</a>. If you hit issues on Windows specifically, open an issue — that’s the platform I’ve tested least and I’d rather know than not.</p><p>First Medium article. It went longer than I planned. That feels about right for a first one.</p><p><em>Vamshi Pokala is a Principal AI Engineer currently based in Boston. He writes about applied AI, architecture, and the gap between what tools promise and what they actually do.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=52f7cbdd051f" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>