<?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 Ana Milutinovic on Medium]]></title>
        <description><![CDATA[Stories by Ana Milutinovic on Medium]]></description>
        <link>https://medium.com/@ana.dip?source=rss-bb3555335f54------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*LpiSlZ-aMeFgHFteulEARQ.png</url>
            <title>Stories by Ana Milutinovic on Medium</title>
            <link>https://medium.com/@ana.dip?source=rss-bb3555335f54------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 18:34:07 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ana.dip/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[A week in Superset, a day in MotherDuck Dives]]></title>
            <link>https://medium.com/arconsis/a-week-in-superset-a-day-in-motherduck-dives-c9fd3f1f3267?source=rss-bb3555335f54------2</link>
            <guid isPermaLink="false">https://medium.com/p/c9fd3f1f3267</guid>
            <category><![CDATA[motherduck]]></category>
            <category><![CDATA[dive]]></category>
            <category><![CDATA[apache-superset]]></category>
            <category><![CDATA[bi-dashboards]]></category>
            <category><![CDATA[data-storytelling]]></category>
            <dc:creator><![CDATA[Ana Milutinovic]]></dc:creator>
            <pubDate>Thu, 05 Nov 2026 12:26:51 GMT</pubDate>
            <atom:updated>2026-11-05T12:26:51.485Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*E3ULsHLmuzm0o8KB86Mv7g.png" /></figure><p>I recently built the same kind of dashboard twice, once as custom Apache Superset plugins and once as a MotherDuck Dive. This article compares the effort and shows how I built the Dive. The numbers come from one test, so read them as my results, not as a benchmark.</p><p><strong>TL;DR</strong></p><ul><li>Superset needed five custom plugins, 30+ prompt revisions, 8+ hours to a first working plugin, and a Kubernetes redeployment for every change.</li><li>The Dive needed 2 prompt revisions and half an hour to a first working version. The finished dashboard took about a day, including learning the tool.</li><li>The result depends on the data: mine was a golden layer, one flat table with joins and metrics already done. Multiple raw tables are untested.</li><li>For customer proposals, a Dive on anonymized data can turn an idea into a clickable demo within a day.</li></ul><h3>Where I’m coming from</h3><p>I’m a Data Visualisation Specialist at <a href="https://www.arconsis.com/">arconsis</a>. I build dashboards for customers, in the tools they already have, the tools they ask for, or the tool I believe fits their needs best.</p><p>I started in Tableau and have also worked with Power BI and GoodData. I tested Metabase, Lightdash, and Fastero. For an internal project, I built custom React visualization plugins for Apache Superset, because the existing charts did not cover what we needed.</p><p>I found MotherDuck by accident. I wanted a quick way to load a test CSV and explore it in Lightdash. Then I saw Dives and the prompt library behind them (<a href="https://motherduck.com/dive-gallery/">MotherDuck Dives</a>). My first reaction was “what the heck?” A few days later I tried it.</p><p>Others have already tested MotherDuck’s speed. Artefact’s benchmark, for example, compares query and dashboard performance on Superset against BigQuery and PostgreSQL (<a href="https://medium.com/artefact-engineering-and-data-science/motherduck-explained-how-it-fits-into-your-data-stack-with-benchmarks-e41c308ebbd4">MotherDuck Explained</a>). Artefact is a MotherDuck launch partner in Europe, so read it as partner content. I looked at something different: how much effort it takes to build the dashboard.</p><h3>The starting point</h3><p>The data was one flat table: a CSV exported from PostgreSQL, with all joins and metrics already calculated. In other words, a golden layer, which is the best case for an agent. This matters for the comparison. The agent did not have to find join paths or decide how a metric is defined.</p><h3>What Dives are</h3><p>Dives are interactive visualizations you create with natural language, directly on top of your data in MotherDuck. You ask an AI agent, and it generates a persistent component that lives in your workspace next to your SQL.</p><p>Technically, a Dive is a React component. It uses a hook called useSQLQuery to run SQL against your MotherDuck databases, Recharts for the charts, and Tailwind CSS for styling. The component queries live data, so a reload always shows current numbers.</p><p>That is the same React and SQL stack I already use for Superset plugins, without the plugin plumbing.</p><h3>The Superset way: one week</h3><p><strong>The goal:</strong> two tables with four-level dropdowns, plus a pivot table where users add columns interactively and pick them from dropdowns.</p><p><strong>Why custom plugins:</strong> the existing Superset charts did not cover three things:</p><ul><li><strong>Drill-down:</strong> the existing table options did not offer it.</li><li><strong>A flexible pivot:</strong> the existing pivot table did not let users freely add columns and rows, so they could not adjust the report on the go.</li><li><strong>A branded header and footer:</strong> I wanted a page with the company branding, showing the customer name and the date range selected in the filters. A chart that pulls both from a dataset that includes the company branding is not something the existing charts offer, so I built one.</li></ul><p><strong>What it took:</strong> five custom React plugins, built with five agents:</p><ul><li><strong>Analytics:</strong> clarifies the data and the questions the dashboard must answer</li><li><strong>Design:</strong> knows all of the company’s design rules and constraints, from logo usage to colors, typography, and tone</li><li><strong>UI:</strong> writes the React plugin</li><li><strong>Grill:</strong> challenges the result and looks for weaknesses</li><li><strong>Judge:</strong> decides whether it is ready to ship</li></ul><p><strong>The numbers:</strong></p><ul><li>30+ revisions of the Superset prompt</li><li>8+ hours until the first plugin ran</li><li>one week until the tables and the pivot were usable</li></ul><p>After that came the manual work:</p><ul><li>add the plugin to the dashboard by hand</li><li>set up every filter</li><li>find a workaround for the date filter</li><li>apply the company design and logo in Superset itself</li><li>choose the right chart type</li></ul><p>On production, every change and every new plugin meant redeploying Superset on Kubernetes. That is part of working with open source: you control everything, and you also run everything. I’m not complaining, I’m showing what it takes to build something like this.</p><p><strong>Why it matters:</strong> each change cost a round of build, deploy, and manual testing. For a small request like an additional button, that is a lot of overhead.</p><h3>The Dives way: two revisions</h3><h4>Setup</h4><p>Connect an AI client to the MotherDuck MCP server. With Claude Code, it is one command:</p><pre>claude mcp add MotherDuck --transport http https://api.motherduck.com/mcp</pre><p>Ask explicitly for a Dive, for example “create a Dive” or “save this as a Dive in MotherDuck”. Otherwise the agent may create a chart without saving it as one.</p><h4>The prompt</h4><p>I wrote one prompt in plain English, using a sample prompt with eight sections as the structure.</p><p>The design rules I had given a whole agent in Superset became two sections in this prompt: Visual design and Wording.</p><p>Here is the full sample prompt. Copy it, fill in what applies, and skip the rest.</p><pre>DIVE: &lt;name&gt;<br>DATA<br>- Source: &lt;database.schema.table, or several tables and how they join&gt;<br>- One row is: &lt;what a single record represents&gt;<br>- Key columns:<br>  - &lt;column&gt;: &lt;meaning, unit, allowed values&gt;<br>  - &lt;value with special meaning&gt;: &lt;how to treat it&gt;<br>- Leave out: &lt;columns, values or records to ignore&gt;<br>- Sensitive data: &lt;fields to hide, aggregate or never show&gt;<br>- Data quirks: &lt;known gaps, duplicates, odd formats, update frequency&gt;</pre><pre>AUDIENCE AND PURPOSE<br>- Who uses it: &lt;role, device, how often&gt;<br>- Questions it must answer:<br>  - &lt;question 1&gt;<br>  - &lt;question 2&gt;<br>  - &lt;question 3&gt;<br>- Level of detail: &lt;aggregate only | individual records allowed&gt;</pre><pre>STRUCTURE<br>- Tabs (about 3, split bigger ideas into separate Dives):<br>  - &lt;tab 1&gt;: &lt;metrics, charts, tables&gt;<br>  - &lt;tab 2&gt;: &lt;metrics, charts, tables&gt;<br>- Default state: &lt;time range, preselected filters, landing tab&gt;</pre><pre>FILTERS AND INTERACTION<br>- Filters: &lt;dimensions, single or multi-select, with search&gt;<br>- Click behavior: &lt;drill down or cross-filter, what happens&gt;<br>- Filters in shared link: &lt;yes | no&gt;<br>- Other: &lt;search, sorting, keyboard shortcuts, detail panels&gt;</pre><pre>VISUAL DESIGN<br>- Theme: &lt;light | dark | both | special style&gt;<br>- Colors (hex): primary &lt;...&gt;, accent &lt;...&gt;, neutrals &lt;...&gt;<br>- Fonts: &lt;...&gt;<br>- Icons: &lt;style, e.g. line icons only&gt;<br>- Logo: &lt;file or none&gt;<br>- Accessibility: &lt;e.g. contrast at least 4.5:1, respect reduced motion&gt;</pre><pre>WORDING<br>- Tone: &lt;e.g. neutral, friendly, formal&gt;<br>- Rules: &lt;style rules, terms to use or avoid&gt;<br>- Empty and loading states: &lt;what they should say&gt;</pre><pre>EXPORTS<br>- Format: &lt;format, or none&gt;<br>- File naming: &lt;e.g. include active filters&gt;</pre><pre>OUT OF SCOPE<br>- &lt;what not to build or show&gt;</pre><p>The sample prompt does the thinking up front. The more the agent knows about the people and the questions, the more it gets right on its own.</p><h4>The result</h4><p>The Dive had three tabs, working filters, and the fun facts and notes in the style I defined under Wording.</p><p>It also had something I never asked for: a clickable pop-up with the details behind each number. The prompt described the audience and their questions, and the Dive added drill-down on its own. That is the request I have heard in every BI tool I have used, and here it was already there.</p><p><strong>The numbers:</strong></p><ul><li>2 revisions of the prompt</li><li>0.5 hours until the first working version</li><li>about a day until the finished dashboard, including learning the tool</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BS71CFB8ZaJnKHAd-ow_ug.png" /><figcaption>Mission Control</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ly4FGOwmXd7Iy-SJqrYP3Q.png" /></figure><h4>Adjusting the code</h4><p>A Dive is plain React and SQL, so I could open it and change it. There is also a Git-based workflow for developing, previewing, and deploying Dives as code. In my test, a change showed up as soon as I saved it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aDBDZ51H8ew1xWu4C0Nebg.png" /></figure><h4>Restyling</h4><p>For fun, I asked for a terminal theme. I wrote a detailed prompt, but a sentence like “make it terminal style” also works. It took around 3 minutes. The prompt changed only the look and wording. The data and structure stayed the same.</p><p>My prompt uses the same structure as the sample above:</p><pre>THEME: Terminal mode for this Dive<br>TASK<br>- Add a theme toggle with two options: &quot;Standard&quot; and &quot;Terminal mode&quot;<br>- Store the selected theme in useDiveState<br>- Keep data and filters the same, change only the look and wording</pre><pre>DATA CHECK<br>- Inspect the Dive&#39;s data and identify:<br>  - measure: &lt;main metric, e.g. hours, revenue, orders&gt;<br>  - dimensions: &lt;main filters, e.g. customer, region, product&gt;<br>  - date: &lt;date column&gt;<br>  - detail: &lt;detail or text column, if any&gt;<br>- Use the real column names in lowercase snake_case</pre><pre>VISUAL DESIGN<br>- Fonts: JetBrains Mono or IBM Plex Mono (Google Fonts), everywhere<br>- Colors (hex): background #000000, primary text &lt;default #5e7dff&gt;,<br>  secondary text #88898c, highlights #FFFFFF<br>- Accent: &lt;default #fc3e46&gt;, used for one highlight badge only<br>- Effects: subtle CRT scanlines, slight text glow, blinking block<br>  cursor after the last line of each panel<br>- Panels: terminal windows with a thin border and a title bar<br>  e.g. &quot;● ● ●  user@&lt;dive-name&gt;:~/&lt;measure&gt;&quot;<br>- Shape: no rounded corners, no shadows, 1px borders #404044</pre><pre>BOOT SEQUENCE<br>- Show on first load for about 1.5 seconds, skippable on click:<br>    &gt; booting &lt;dive-name&gt; v1.0 ...<br>    &gt; connecting to motherduck ........ [ OK ]<br>    &gt; loading &lt;table&gt; ................. [ OK ] &lt;row count&gt; rows<br>    &gt; agents online ................... [ OK ]<br>    &gt; welcome. &lt;optional tagline&gt;</pre><pre>FILTERS AS COMMANDS<br>- Show active filters as a live command line that updates with the dropdowns:<br>    $ &lt;measure&gt; --&lt;dimension_1&gt;=&quot;Value A&quot; --&lt;date&gt;=2026-01..2026-06<br>- Allow typing: one flag per filter, applied on enter<br>- Unknown flags: friendly &quot;command not found&quot;, hint &quot;try: help&quot;<br>- help: print the available flags<br>- clear: reset all filters</pre><pre>OVERVIEW<br>- KPIs as aligned key/value lines:<br>    &lt;measure&gt;_total ..... 1,240.0<br>    records ............. 3,412<br>- ASCII bar charts for the top 2 dimensions:<br>    Value A  ████████████████░░░░  42%  520.5<br>- Summary as a log line:<br>    [INFO] &lt;period&gt; · &lt;total&gt; &lt;measure&gt; across &lt;n&gt; &lt;dimension_1&gt;s<br>- Fun translations as [FUN] lines, if they fit the measure<br>  e.g. hours to working weeks, revenue to coffees</pre><pre>LOGS<br>- Detail rows as log lines, newest at the bottom, auto-scroll:<br>    &lt;date&gt;  &lt;dimension_1&gt;  &lt;dimension_2&gt;  &lt;measure&gt;  &quot;&lt;detail&gt;&quot;<br>- Search box styled as &quot;grep: ▌&quot;, matches highlighted<br>- Click on a line: expand the full record</pre><pre>TABLES AND PIVOTS<br>- Section title as a command e.g. &quot;$ pivot&quot;<br>- Controls as CLI flags: --rows=... --cols=... --values=...<br>- Monospace table with aligned columns, totals row separated by &quot;────&quot;<br>- Download button labelled &quot;$ export --format=csv&quot;</pre><pre>EASTER EGGS<br>- sudo coffee: prints &quot;☕ permission granted.&quot;<br>- whoami: prints &quot;operator of &lt;dive-name&gt;&quot;<br>- matrix: brief falling characters in the primary color, then back to normal</pre><pre>EMPTY STATES<br>- No results: &quot;[WARN] 0 results for this filter.&quot;<br>- No search match: &quot;grep: no matches. try a broader search.&quot;</pre><pre>WORDING<br>- Tone: short, neutral, slightly informal<br>- Rules: aggregates only, no rankings of individual people</pre><pre>ACCESSIBILITY<br>- Contrast at least 4.5:1<br>- Respect prefers-reduced-motion: skip boot animation, glow and matrix effect</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LEiYPsbkfQ-Rl7b-dfAMJg.png" /><figcaption>Terminal Style Mission Control</figcaption></figure><p>In Superset, a restyle means CSS work or a rebuild. Here it was one prompt.</p><h4>A fair comparison</h4><p>These numbers are not a lab test. Superset needed custom plugins and a deployment pipeline. The Dive came with a sample prompt and a prepared data layer. The Superset week covered two drill-down tables, an interactive pivot, and a branded header and footer. Both are honest timelines of what I actually spent.</p><h3>Why a golden layer matters</h3><p>Everything I read pointed in the same direction: Dives work best on a golden layer, meaning data that is already cleaned, joined, and defined before the agent sees it. My own result fits that. With one flat table where joins and metrics were already done, the first working version took half an hour, and the quality surprised me.</p><p>The reason is simple. The agent does not have to guess join paths or decide how a metric is defined, so there is less room for wrong numbers. MotherDuck’s own tests on a small question set point the same way: agents with good context make far fewer query errors. That is the vendor’s benchmark, not independent proof, so treat it as a hint.</p><p><strong>What this means in practice:</strong> put the effort into the data layer first, and the prompt second.</p><h3>What my test did not cover</h3><p>My data was a single flat CSV, uploaded by hand. That is the easy case for an agent.</p><p><strong>Flights.</strong> Flights are scheduled Python jobs that run inside MotherDuck. An agent connected to the MCP server can write the pipeline, create the Flight, and schedule it. I did not need them for this test. If the data has to refresh on a schedule, that is where Flights would come in. They are in public preview, so check their status first.</p><p><strong>Guides and the semantic layer.</strong> MotherDuck’s answer to the semantic layer question is Guides: versioned markdown documents stored in the warehouse that give agents business context, such as what a column means and how tables relate. With a pre-joined table, I had no need for them. How a Dive behaves across several raw tables is still untested.</p><p><strong>Why it matters:</strong> a golden layer made my result possible. Multiple raw tables will be the real test, and I plan to try that next.</p><h3>Where Dives fit, and where they don’t</h3><ul><li><strong>Cloud only.</strong> MotherDuck is cloud infrastructure. It runs in AWS regions across the US, Europe, and Asia Pacific, including Frankfurt and Dublin in Europe. You choose the region when you create the organization, and you cannot change it afterward. If data must stay on premises, it is not an option.</li><li><strong>Access control.</strong> MotherDuck has role-based access control. I only looked at it briefly, so check the documentation for how it applies to Dives. MotherDuck is SOC 2 Type II and GDPR compliant, with HIPAA BAAs available on request for Business plan customers.</li><li><strong>Cost.</strong> Pricing is pay-as-you-go. The Lite plan has no platform fee and includes 10 compute unit hours and 10 GB of storage per month. It is limited to the smallest instance type and 3 users. The Business plan costs $250 per month plus usage. Dives are available on all plans at no extra charge, while embedding needs Business.</li><li><strong>Workflow.</strong> You switch between your AI client and the browser, which does not match the otherwise integrated workspace.</li><li><strong>Verification.</strong> A generated dashboard can still be wrong. Check the SQL behind every number.</li></ul><h3>When I would still pick Superset</h3><p>Superset is open source, self-hosted, and gives you full control over code, deployment, and where your data lives. I would still choose it when data has to stay on premises, when a project needs deep customization that you want to own, or when open source is a requirement. I would reach for a Dive when I want a fast, good-looking dashboard on a prepared data layer in the cloud.</p><h3>Where this helps at arconsis: demos and proposals</h3><p>Many of our customers work in regulated industries and have strict rules about their data. That limits what we can show them before a project starts. With anonymized or synthetic data, a Dive turns an idea into a working, branded dashboard within a day. In a proposal or a presentation, customers can click through an idea instead of reading a description.</p><p>Data location is one of the first questions customers ask. If you choose a European region when you create the organization, the data stays in the EU. Two cautions: the AI client you connect also sees your prompts and query results, so check that setup with the same care, and for demos I use anonymized data only.</p><h3>Why this matters</h3><p>Generated output only helps if you can check it. A dashboard still needs correct numbers, traceable logic, and controlled access. Dives give you the SQL and the code, so a review is possible. Whether the setup is enough for your data depends on your requirements, so test it with a small project first.</p><h3>Takeaway</h3><p>Superset and Tableau are not going away, and I still build custom plugins when a project needs full control. But the work that cost me a week, wiring tables, pivots, filters, and drill-downs, took me about a day in Dives, with two prompt revisions, including learning the tool. The result depends heavily on the data layer: with a golden layer behind it, a Dive worked very well for me.</p><p>The right tool depends on the customer: what they already run, where their data must live, and how much control they need. A Dive is one more option I can offer for quick, good-looking dashboards on a prepared data layer, and a fast way to turn an idea into something a customer can click through before a project starts. That leaves more time for the part that matters: checking that the numbers are right.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c9fd3f1f3267" width="1" height="1" alt=""><hr><p><a href="https://medium.com/arconsis/a-week-in-superset-a-day-in-motherduck-dives-c9fd3f1f3267">A week in Superset, a day in MotherDuck Dives</a> was originally published in <a href="https://medium.com/arconsis">arconsis</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>