<?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 André Bergholz on Medium]]></title>
        <description><![CDATA[Stories by André Bergholz on Medium]]></description>
        <link>https://medium.com/@bergholz?source=rss-f3a56a39aa4f------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*Ivy7TU4bMOe8Mxkw.jpg</url>
            <title>Stories by André Bergholz on Medium</title>
            <link>https://medium.com/@bergholz?source=rss-f3a56a39aa4f------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 16:39:59 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@bergholz/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[An LLM Wiki for the Rest of Us]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://aiadvances.org/an-llm-wiki-for-the-rest-of-us-2a2b19fe7dd3?source=rss-f3a56a39aa4f------2"><img src="https://cdn-images-1.medium.com/max/1672/1*_ZyfuxPU4NrSZDa4SeEpag.png" width="1672"></a></p><p class="medium-feed-snippet">Keep the wiki your team edits. Let the model compile everything else.</p><p class="medium-feed-link"><a href="https://aiadvances.org/an-llm-wiki-for-the-rest-of-us-2a2b19fe7dd3?source=rss-f3a56a39aa4f------2">Continue reading on AI Advances »</a></p></div>]]></description>
            <link>https://aiadvances.org/an-llm-wiki-for-the-rest-of-us-2a2b19fe7dd3?source=rss-f3a56a39aa4f------2</link>
            <guid isPermaLink="false">https://medium.com/p/2a2b19fe7dd3</guid>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[obsidian]]></category>
            <category><![CDATA[knowledge-management]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[claude-code]]></category>
            <dc:creator><![CDATA[André Bergholz]]></dc:creator>
            <pubDate>Tue, 10 Nov 2026 16:03:02 GMT</pubDate>
            <atom:updated>2026-11-10T16:03:02.707Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The AI Race Has No Finish Line]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://generativeai.pub/the-ai-race-has-no-finish-line-b3ce8f882dbd?source=rss-f3a56a39aa4f------2"><img src="https://cdn-images-1.medium.com/max/1774/1*4rNEetEnINkznm6yWOilQA.png" width="1774"></a></p><p class="medium-feed-snippet">The race is a Moloch, a prisoner&#x2019;s dilemma that never ends</p><p class="medium-feed-link"><a href="https://generativeai.pub/the-ai-race-has-no-finish-line-b3ce8f882dbd?source=rss-f3a56a39aa4f------2">Continue reading on Generative AI »</a></p></div>]]></description>
            <link>https://generativeai.pub/the-ai-race-has-no-finish-line-b3ce8f882dbd?source=rss-f3a56a39aa4f------2</link>
            <guid isPermaLink="false">https://medium.com/p/b3ce8f882dbd</guid>
            <category><![CDATA[ai-safety]]></category>
            <category><![CDATA[game-theory]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[anthropic-claude]]></category>
            <category><![CDATA[openai]]></category>
            <dc:creator><![CDATA[André Bergholz]]></dc:creator>
            <pubDate>Tue, 29 Sep 2026 12:53:28 GMT</pubDate>
            <atom:updated>2026-10-01T08:55:55.477Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[API for Volume, MCP for Chat, CLI for Terminals]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://pub.towardsai.net/api-for-volume-mcp-for-chat-cli-for-terminals-2785cd6a9efc?source=rss-f3a56a39aa4f------2"><img src="https://cdn-images-1.medium.com/max/2172/1*35v6wXzuTp8g6uCqPe7glg.png" width="2172"></a></p><p class="medium-feed-snippet">Five common jobs, their corresponding commands, and the two checks that differentiate them</p><p class="medium-feed-link"><a href="https://pub.towardsai.net/api-for-volume-mcp-for-chat-cli-for-terminals-2785cd6a9efc?source=rss-f3a56a39aa4f------2">Continue reading on Towards AI »</a></p></div>]]></description>
            <link>https://pub.towardsai.net/api-for-volume-mcp-for-chat-cli-for-terminals-2785cd6a9efc?source=rss-f3a56a39aa4f------2</link>
            <guid isPermaLink="false">https://medium.com/p/2785cd6a9efc</guid>
            <category><![CDATA[api]]></category>
            <category><![CDATA[model-context-protocol]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[André Bergholz]]></dc:creator>
            <pubDate>Thu, 17 Sep 2026 19:31:02 GMT</pubDate>
            <atom:updated>2026-09-17T20:17:38.131Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Probably, Approximately, Shipped]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://aiadvances.org/probably-approximately-shipped-4c7e4366a6a6?source=rss-f3a56a39aa4f------2"><img src="https://cdn-images-1.medium.com/max/1774/1*iaNoKlt4LudEm8GOBVUGjQ.png" width="1774"></a></p><p class="medium-feed-snippet">Building software on a model that changes every six weeks breaks the one assumption the whole discipline was built on.</p><p class="medium-feed-link"><a href="https://aiadvances.org/probably-approximately-shipped-4c7e4366a6a6?source=rss-f3a56a39aa4f------2">Continue reading on AI Advances »</a></p></div>]]></description>
            <link>https://aiadvances.org/probably-approximately-shipped-4c7e4366a6a6?source=rss-f3a56a39aa4f------2</link>
            <guid isPermaLink="false">https://medium.com/p/4c7e4366a6a6</guid>
            <category><![CDATA[claude-code]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[prompt-engineering]]></category>
            <dc:creator><![CDATA[André Bergholz]]></dc:creator>
            <pubDate>Wed, 02 Sep 2026 18:37:50 GMT</pubDate>
            <atom:updated>2026-09-02T18:37:50.175Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Too Smart to Be Right: Why Fluent Language Isn’t a Sign of Good Reasoning]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://generativeai.pub/too-smart-to-be-right-why-fluent-language-isnt-a-sign-of-good-reasoning-424cf69d621a?source=rss-f3a56a39aa4f------2"><img src="https://cdn-images-1.medium.com/max/1536/1*TeKrU0crje-y2VmHUeum6Q.png" width="1536"></a></p><p class="medium-feed-snippet">A confident, well-phrased answer earns trust it hasn&#x2019;t earned. From a machine, and from a person.</p><p class="medium-feed-link"><a href="https://generativeai.pub/too-smart-to-be-right-why-fluent-language-isnt-a-sign-of-good-reasoning-424cf69d621a?source=rss-f3a56a39aa4f------2">Continue reading on Generative AI »</a></p></div>]]></description>
            <link>https://generativeai.pub/too-smart-to-be-right-why-fluent-language-isnt-a-sign-of-good-reasoning-424cf69d621a?source=rss-f3a56a39aa4f------2</link>
            <guid isPermaLink="false">https://medium.com/p/424cf69d621a</guid>
            <category><![CDATA[large-language-models]]></category>
            <category><![CDATA[critical-thinking]]></category>
            <category><![CDATA[cognitive-science]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[André Bergholz]]></dc:creator>
            <pubDate>Sun, 16 Aug 2026 02:19:49 GMT</pubDate>
            <atom:updated>2026-08-16T20:55:33.946Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The Turing Award’s AI Story]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://aiadvances.org/the-turing-awards-ai-story-f79b12ab7699?source=rss-f3a56a39aa4f------2"><img src="https://cdn-images-1.medium.com/max/1086/1*-hyL2kpLjX4smkg1-w97zA.png" width="1086"></a></p><p class="medium-feed-snippet">What computer science&#x2019;s highest prize reveals about who shaped AI and what its silences leave unspoken</p><p class="medium-feed-link"><a href="https://aiadvances.org/the-turing-awards-ai-story-f79b12ab7699?source=rss-f3a56a39aa4f------2">Continue reading on AI Advances »</a></p></div>]]></description>
            <link>https://aiadvances.org/the-turing-awards-ai-story-f79b12ab7699?source=rss-f3a56a39aa4f------2</link>
            <guid isPermaLink="false">https://medium.com/p/f79b12ab7699</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[turing-award]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[history]]></category>
            <dc:creator><![CDATA[André Bergholz]]></dc:creator>
            <pubDate>Mon, 13 Jul 2026 22:01:44 GMT</pubDate>
            <atom:updated>2026-07-13T22:01:44.210Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[My AI Workflow, One Year Later]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://levelup.gitconnected.com/my-ai-workflow-one-year-later-5ebfba2934dc?source=rss-f3a56a39aa4f------2"><img src="https://cdn-images-1.medium.com/max/1376/1*Q8gAufMvqsoyuVglmUKS5A.png" width="1376"></a></p><p class="medium-feed-snippet">A year ago I said this would feel outdated in six months. I was right.</p><p class="medium-feed-link"><a href="https://levelup.gitconnected.com/my-ai-workflow-one-year-later-5ebfba2934dc?source=rss-f3a56a39aa4f------2">Continue reading on Level Up Coding »</a></p></div>]]></description>
            <link>https://levelup.gitconnected.com/my-ai-workflow-one-year-later-5ebfba2934dc?source=rss-f3a56a39aa4f------2</link>
            <guid isPermaLink="false">https://medium.com/p/5ebfba2934dc</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[future-of-work]]></category>
            <dc:creator><![CDATA[André Bergholz]]></dc:creator>
            <pubDate>Wed, 24 Jun 2026 16:32:57 GMT</pubDate>
            <atom:updated>2026-06-24T16:32:57.097Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Pre-Training Gives LLMs Their Capability. Post-Training Gives Them Their Behavior.]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://generativeai.pub/pre-training-gives-llms-their-capability-post-training-gives-them-their-behavior-e75f7039a2b2?source=rss-f3a56a39aa4f------2"><img src="https://cdn-images-1.medium.com/max/2600/1*iEDbhdbS3JrTftpbxJezXA.png" width="2848"></a></p><p class="medium-feed-snippet">The model you use has two biographies&#x200A;&#x2014;&#x200A;and one of them you can help shape</p><p class="medium-feed-link"><a href="https://generativeai.pub/pre-training-gives-llms-their-capability-post-training-gives-them-their-behavior-e75f7039a2b2?source=rss-f3a56a39aa4f------2">Continue reading on Generative AI »</a></p></div>]]></description>
            <link>https://generativeai.pub/pre-training-gives-llms-their-capability-post-training-gives-them-their-behavior-e75f7039a2b2?source=rss-f3a56a39aa4f------2</link>
            <guid isPermaLink="false">https://medium.com/p/e75f7039a2b2</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[large-language-models]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[fine-tuning]]></category>
            <dc:creator><![CDATA[André Bergholz]]></dc:creator>
            <pubDate>Tue, 02 Jun 2026 08:18:12 GMT</pubDate>
            <atom:updated>2026-06-02T08:18:12.888Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Slides Are a Build Artifact Now]]></title>
            <link>https://ai.plainenglish.io/slides-are-a-build-artifact-now-c3ef24dc6f60?source=rss-f3a56a39aa4f------2</link>
            <guid isPermaLink="false">https://medium.com/p/c3ef24dc6f60</guid>
            <category><![CDATA[presentations]]></category>
            <category><![CDATA[claude-code]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[André Bergholz]]></dc:creator>
            <pubDate>Fri, 22 May 2026 22:10:46 GMT</pubDate>
            <atom:updated>2026-11-10T15:14:03.236Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="A technical blueprint on dark navy paper shows the schematic layout of a presentation slide: labeled dimensions, content zones. Beside it, the finished slide rendered in clean white: the same content now formatted as a polished corporate presentation with typography, hierarchy, and a styled architecture diagram." src="https://cdn-images-1.medium.com/max/1024/1*zYn5aSyR8MYAFeO0XlX9pg.png" /><figcaption>(This image was generated using AI technology.)</figcaption></figure><p><strong>A reproducible pipeline from Markdown to boardroom-ready PDF</strong></p><blockquote>Every company pays the slide tax. Nobody budgets for it.</blockquote><p>Most companies have a PowerPoint template. It has the right colors, the right fonts, the approved layouts. That part is solved. What remains is still manual: deciding which layout fits this particular slide, how many slides the content actually needs, what goes on the title versus the body, whether this point deserves its own slide or belongs to the one before it. Someone makes those decisions every time, for every presentation, one slide at a time.</p><blockquote>I wanted to eliminate not just formatting work, but the recurring layout decisions behind each slide.</blockquote><p>This setup makes most sense for teams that already work close to the command line — developers, technical leads, strategy teams comfortable with plain text. If your workflow centers on real-time collaborative editing in Google Slides, this is probably overkill. If you version-control your documents and want your presentations to follow the same logic, read on.</p><p>Tools like NotebookLM, Gamma or Beautiful.ai take a different approach: describe what you want, get slides in a browser, done. For many people that works well. I work with text files and command lines. If you version-control your documents and want your presentations to follow the same logic, this is for you.</p><h4>The Framework: Two Hard Requirements</h4><p>I had two constraints before I looked at any tooling. The output had to be a real PDF: correct fonts, spacing, and colors every time. And the source had to be plain text, ideally Markdown, so I could version-control it and pass it to a language model without wrestling with a binary format.</p><p>PowerPoint and Google Slides are workable for both, but neither is clean. Their file formats are blobs, PDF export behavior varies across platforms and versions, and neither plays well with version control or automated generation. RevealJS is closer — it has Markdown support built in — but for anything beyond simple single-column slides, you drop into raw HTML. Multi-column layouts, complex content blocks, custom positioning: all HTML.</p><p>Quarto adds a Markdown abstraction layer on top of RevealJS that covers the common cases. Multi-column layouts, incremental lists, code highlighting — all expressible in Markdown syntax. The custom components in the template still use HTML and CSS, but that code lives in the template, not in the presentation content. Day-to-day authoring stays in plain text. PDF export is native to RevealJS and works without configuration.</p><h4>The Template: The Step That Makes Everything Else Work</h4><p>I set up a Claude Code project with a read-only context directory containing the corporate identity. Not a prose description of the brand, but the actual values. Primary color, accent and signal colors, text tones, the gradient used across branded materials, the font stack with all its weights. Token names mapped directly to SCSS variables.</p><p>Not everything from the CI went in. Full color scales, shadow variables, border-radius tokens were left out. The presentation layer doesn’t need them, and an overloaded context produces inconsistent output.</p><p>The read-only constraint is deliberate, and it’s actually standard practice in almost all my Claude Code projects. It functions as a lightweight alternative to RAG: the model gets access to reference material it can draw from but never modify. No infrastructure required, just a directory the AI knows it cannot touch. Here, the CI directory plays that role. The brand definition stays fixed; generation references it.</p><p>The prompt to Claude was specific: generate a component library for business presentations, covering every common slide scenario, built from these brand tokens and constrained to what a typical corporate deck needs.</p><p>The output was a component system: structural slide types for titles, section dividers, and CTAs; multi-column layouts; content components for stats, timelines, before/after comparisons, team grids, checklists, and more; callout box variants; image treatments; background variants. Each component a named CSS class or HTML block, all in the correct colors and fonts. The kind of vocabulary that covers whatever a business presentation needs to say.</p><h4>Docker Compose: From Personal Tool to Team Infrastructure</h4><blockquote>A template without a reproducible build is fragile.</blockquote><p>It works on the machine where it was built. It breaks when someone else clones the repo, when the local Quarto version drifts, or when a required font isn’t installed.</p><p>I packaged the build in Docker Compose: Quarto version pinned, fonts installed in the container, compilation scripted. Two services: one that serves the presentation as a browser-based slideshow, one that exports the PDF. docker-compose up for the live preview, docker-compose up pdf for the export. Anyone on the team gets the same output regardless of local setup. That&#39;s what makes this a team tool rather than a personal experiment.</p><h4>Creating an Actual Presentation</h4><blockquote>For each new presentation, I create a fresh Claude Code project.</blockquote><p>The pattern repeats: the template repository goes in as a read-only context directory, the same role the CI played one level up. Reference material the model draws from but cannot change. A second directory holds the content: the topic, how I want the slides structured, key points per slide, any data or quotes I want to include.</p><blockquote>One thing matters in the prompt: be explicit that Claude should only use components that exist in the template.</blockquote><p>Without that constraint, it sometimes invents its own layouts. They work often, but equally often they don’t, and the output is no longer strictly on-brand. The prompt needs to make clear that the template is the boundary, not a suggestion.</p><p>In practice the content directory might contain something like:</p><pre>Q2 results: revenue up 12%, churn down significantly<br>Key drivers: enterprise expansion, reduced time-to-value, new fintech vertical<br>Team: two new AEs, one CSM added in Q2</pre><p>Claude decides how many slides that needs, which component fits each piece (e.g., a stat block for the numbers, a three-column layout for the drivers, a team grid for the hires) and generates the Markdown. docker-compose up to preview in the browser, docker-compose up pdf to export.</p><p>The fonts are correct because they’re installed in the container. The colors are correct because they came from the CI directory. The layouts hold because the template defines them. Writing the content is the only work left.</p><h4>Where This Doesn’t Fit</h4><p>A few cases where this setup is the wrong tool.</p><p>If your presentations require highly custom visuals per slide — bespoke illustrations, complex data visualizations, layouts that differ significantly from slide to slide — the component system will feel constraining. It’s designed for consistency, not creative variety.</p><p>If your team needs real-time collaborative editing, browser-based tools like Google Slides or Gamma handle that better. This pipeline produces files, not live documents.</p><p>And if the investment in template setup isn’t justified by volume — say, you produce two or three presentations a year — the overhead probably doesn’t pay off. The break-even point is somewhere around regular cadence: weekly decks, recurring reports, standardized pitch materials.</p><h4>What This Actually Changes</h4><p>The time savings are real, but they’re not the main point.</p><p>When a template exists as a versioned, machine-readable artifact and the build is reproducible, the PowerPoint file stops being the deliverable.</p><blockquote>The slide deck becomes derived output: the same relationship as source code to a compiled binary.</blockquote><p>The content is the source. The presentation is what the build produces from it.</p><p>Most corporate workflows run this backward. Every PowerPoint carries its own formatting, its own consistency risks, its own maintenance overhead. That cost compounds across every presentation, every update, every new team member who has to learn the template by hand. Moving design to infrastructure means paying that cost once.</p><p>You encode your brand once. Every presentation after that is just writing.</p><p><em>I write about AI for leaders and professionals who want to understand what’s actually changing. Follow along or explore my full catalog in the </em><a href="https://medium.com/@bergholz/9a5267d85403"><strong><em>AI Writing Compass</em></strong></a>.</p><h3>A message from our Founder</h3><p>Hey, <a href="https://linkedin.com/in/sunilsandhu">Sunil</a> here. I wanted to take a moment to thank you for reading until the end and for being a part of this community. Did you know that our team run these publications as a volunteer effort to over 3.5m monthly readers? We don’t receive any funding, we do this to support the community.</p><p>If you want to show some love, please take a moment to follow me on <a href="https://linkedin.com/in/sunilsandhu">LinkedIn</a>, <a href="https://tiktok.com/@messyfounder">TikTok</a>, <a href="https://instagram.com/sunilsandhu">Instagram</a>. You can also subscribe to our <a href="https://newsletter.plainenglish.io/">weekly newsletter</a>. And before you go, don’t forget to clap and follow the writer️!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c3ef24dc6f60" width="1" height="1" alt=""><hr><p><a href="https://ai.plainenglish.io/slides-are-a-build-artifact-now-c3ef24dc6f60">Slides Are a Build Artifact Now</a> was originally published in <a href="https://ai.plainenglish.io/">Artificial Intelligence in Plain English</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[No Intelligence Without Illusion]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://generativeai.pub/no-intelligence-without-illusion-0e1cc3a58d01?source=rss-f3a56a39aa4f------2"><img src="https://cdn-images-1.medium.com/max/2048/1*1JFBzZ2-boaVE3zdLeTeOw.png" width="2048"></a></p><p class="medium-feed-snippet">The same cognitive architecture that produces human intelligence also produces human delusion. We are now scaling both.</p><p class="medium-feed-link"><a href="https://generativeai.pub/no-intelligence-without-illusion-0e1cc3a58d01?source=rss-f3a56a39aa4f------2">Continue reading on Generative AI »</a></p></div>]]></description>
            <link>https://generativeai.pub/no-intelligence-without-illusion-0e1cc3a58d01?source=rss-f3a56a39aa4f------2</link>
            <guid isPermaLink="false">https://medium.com/p/0e1cc3a58d01</guid>
            <category><![CDATA[ai-safety]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[large-language-models]]></category>
            <category><![CDATA[philosophy]]></category>
            <category><![CDATA[cognitive-science]]></category>
            <dc:creator><![CDATA[André Bergholz]]></dc:creator>
            <pubDate>Mon, 27 Apr 2026 09:31:30 GMT</pubDate>
            <atom:updated>2026-04-27T10:48:04.195Z</atom:updated>
        </item>
    </channel>
</rss>