<?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 Kevin Granger on Medium]]></title>
        <description><![CDATA[Stories by Kevin Granger on Medium]]></description>
        <link>https://medium.com/@Kevin_Granger?source=rss-ba1bcb47af9c------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*KZ_ODVFbs3AGEDsmkgYALQ.png</url>
            <title>Stories by Kevin Granger on Medium</title>
            <link>https://medium.com/@Kevin_Granger?source=rss-ba1bcb47af9c------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 02:27:42 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@Kevin_Granger/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[AI-Native Data, Without the Enterprise Nonsense]]></title>
            <link>https://medium.com/@Kevin_Granger/ai-native-data-without-the-enterprise-nonsense-9f9eb974e25d?source=rss-ba1bcb47af9c------2</link>
            <guid isPermaLink="false">https://medium.com/p/9f9eb974e25d</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[startup]]></category>
            <dc:creator><![CDATA[Kevin Granger]]></dc:creator>
            <pubDate>Thu, 05 Feb 2026 15:26:26 GMT</pubDate>
            <atom:updated>2026-02-05T15:26:26.818Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*icTi7bA1_mjqGhziJtJkNg.png" /><figcaption>Altertable recently opened early access to an AI-native data platform</figcaption></figure><p>The data platform market is worth billions. Recent market research puts global data analytics spending in the tens of billions of dollars, and growing fast (<a href="https://www.statista.com/topics/1464/big-data/">Statista analytics topic page</a>).</p><p>So why do so many teams dread their tools?</p><p>I’m Kevin, co-founder at <a href="https://altertable.ai/">Altertable</a>, and we recently opened early access to an AI-native data platform. But this isn’t a launch post. It’s a memo about usefulness: what it means, why enterprise software lost it, and how an AI-native, agentic approach can finally give people their time back.</p><h3>The enterprise software trap</h3><p>A small team builds something simple and great. People love it. Then enterprise deals arrive, and roadmaps bend toward checklists. Every feature request becomes a must-have. The product turns into a Swiss Army knife that’s dull at everything.</p><p>Docs become a maze. Onboarding needs consultants. What used to take minutes now takes months. The original users quietly leave.</p><p>Data platforms fell into this trap harder than most. The modern data stack brought real power, and with it came tool sprawl, orchestration debt, endless configuration, and brittle pipelines. This pattern shows up again and again in industry retrospectives.</p><p>“Enterprise-ready” slowly became code for “complicated and slow.”</p><h3>What teams actually want</h3><p>Before we wrote a line of code, we spoke with 50+ teams and asked a simple question: <strong>What would your ideal data tool actually do? </strong>The answers weren’t feature lists. They were about time.</p><blockquote>“Let me answer questions without waiting for engineering.”</blockquote><blockquote>“Stop making me configure the same thing three different ways.”</blockquote><blockquote>“Just work with the data I already have.”</blockquote><p>Nobody asked for more dashboards, more integrations, or certifications for their own sake. What they wanted was self-serve analytics that works with their reality. Sometimes that means connecting to an existing warehouse; sometimes it means hosting data directly; often it’s both at once. Same definitions, same metrics, same experience: in minutes, not months.</p><p>Industry writing on self-serve and natural-language analytics echoes this theme: faster answers with less dependency on specialist data teams, enabled by natural language interfaces that turn everyday questions into actionable queries (<a href="https://querio.ai/articles/natural-language-interfaces-data-practical-guide">Natural Language Interfaces for Data: A Practical Guide</a>).</p><h3>Why now: the real AI advantage</h3><p>AI isn’t magic. It’s just very good at one thing software has always struggled with: <strong>understanding context</strong>.</p><p>Most tools force people to learn the system. An AI-native data platform lets the system learn from people instead: from their language, their data, and the way they work.</p><p>That’s where AI agents come in. Instead of a single monolithic “chat with your data” box, you get a set of specialized agents, each responsible for part of the workflow: understanding intent, planning a query, validating access, running analysis, and explaining results.</p><p>This kind of conversational, multi-agent approach is emerging as a new paradigm in analytics, where systems continually negotiate intent, data access, and governance with users.</p><p>Concretely, that enables:</p><ul><li>A <strong>planner agent</strong> that turns a vague question into a concrete analysis plan.</li><li>A <strong>schema and semantics agent</strong> that maps terms like “active customer” or “SMB churn” to your actual definitions.</li><li>A <strong>SQL agent</strong> that proposes queries step by step, which you can inspect and edit before execution.</li><li>A <strong>validation and guardrail agent</strong> that checks permissions and data governance policies.</li><li>An <strong>explanation agent</strong> that translates results into plain language, citing the metrics and segments used.</li></ul><p>The point isn’t to replace humans with one big model; it’s to orchestrate a swarm of narrow, opinionated agents around the way teams already think and work.</p><p>We go deeper on this idea in our essay <a href="https://altertable.ai/blog/2025-06-24-from-dashboards-to-dialogue">From Dashboards to Dialogue</a>.</p><h3>Our principles at Altertable</h3><h3>Speed first: value in minutes, not months</h3><p>Connect to your existing warehouse or let Altertable host the data as a lakehouse. Same model, same definitions, sane defaults, and progressive disclosure as teams need more control.</p><p>Unified lakehouse architectures are gaining traction precisely because they reduce data movement and fragmentation, enabling teams to work from a single source of truth instead of multiple disconnected tools (<a href="https://www.databricks.com/glossary/data-lakehouse">What is a Data Lakehouse?</a>).</p><p>We’ve written more about why speed matters for understanding in data work in <a href="https://altertable.ai/blog">our blog</a>.</p><h3>AI that works</h3><p>Instead of a black-box chatbot, Altertable uses cooperating agents with clear responsibilities. Each step is inspectable (SQL, lineage, reasoning) so humans can review, trust, and learn from the system.</p><h3>Less is more</h3><p>Fewer surfaces, deeper utility. Saying “no” to bloat is how usefulness survives scale. Agents help here too: instead of adding new UI modes for every use case, we add reusable capabilities behind the scenes.</p><h3>What this looks like in practice</h3><p>Imagine asking:</p><p><strong>“What drove churn last quarter among SMB customers?”</strong></p><p>Behind the scenes:</p><ol><li>The intent agent structures the task.</li><li>The semantics agent resolves “churn,” “SMB,” and “customer” to your definitions.</li><li>The planner agent sketches a readable analysis plan.</li><li>The SQL agent generates queries and shows them before execution.</li><li>The validation agent checks permissions and policies.</li><li>The explanation agent summarizes results with links back to SQL and lineage.</li></ol><p>From your point of view, you asked a question and approved a plan. Under the hood, a group of agents handled the boring details.</p><p>When there’s ambiguity (ie:“logo churn or seat churn?”) the system asks a focused follow-up and remembers the answer. Over time, the platform becomes a memory of how your company talks about its data.</p><p>This aligns closely with research on conversational and natural language business intelligence interfaces that reveal how NLQs are translated into executable analytics queries (<a href="https://www.techtarget.com/whatis/definition/natural-language-query">What is Natural Language Query</a>).</p><p>The goal isn’t to replace analysis with a chatbot. It’s to collapse hours of setup, back-and-forth, and rework into a tight feedback loop: <strong>ask, clarify, inspect, ship</strong>.</p><h3>Guardrails by design</h3><p>Usefulness and safety are the same requirement.Every agent operates inside the same permission and policy layer:</p><ul><li>Access respects existing roles and row-level security.</li><li>Sensitive changes are routed through review.</li><li>Lineage, version history, and agent actions remain visible and auditable.</li></ul><p>This mirrors guidance from AI governance bodies that emphasize combining natural-language interfaces with strong policy enforcement and auditability, such as the <a href="https://www.nist.gov/itl/ai-risk-management-framework">NIST AI Risk Management Framework</a> and the <a href="https://oecd.ai/en/ai-principles">OECD AI Principles</a>.</p><h3>The bet we’re making</h3><p>We’re betting people are done with “powerful” tools that require a Ph.D. to use. Simplicity beats feature lists. Faster time to insight matters; without treating security, governance, and compliance as optional.</p><p>And we’re betting that context-aware AI agents can finally deliver intuitive software: less time configuring tools, more time asking better questions.</p><p>If that resonates, come help shape it.</p><p><strong>Early access is open at </strong><a href="https://altertable.ai/"><strong>altertable.ai</strong></a><strong>.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9f9eb974e25d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Emoji fatigue]]></title>
            <link>https://medium.com/@Kevin_Granger/emoji-fatigue-4a07e3806c3f?source=rss-ba1bcb47af9c------2</link>
            <guid isPermaLink="false">https://medium.com/p/4a07e3806c3f</guid>
            <dc:creator><![CDATA[Kevin Granger]]></dc:creator>
            <pubDate>Wed, 07 Dec 2016 14:40:59 GMT</pubDate>
            <atom:updated>2016-12-07T14:40:59.356Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/49/1*Oo0o4M8LpRZ0U741p9y19g.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4a07e3806c3f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A painstakingly crafted search for Hearthstone]]></title>
            <link>https://medium.com/algolia-stories/a-painstakingly-crafted-search-for-hearthstone-c21b3fa4223c?source=rss-ba1bcb47af9c------2</link>
            <guid isPermaLink="false">https://medium.com/p/c21b3fa4223c</guid>
            <category><![CDATA[hearthstone]]></category>
            <category><![CDATA[team-stories]]></category>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[ui]]></category>
            <category><![CDATA[search]]></category>
            <dc:creator><![CDATA[Kevin Granger]]></dc:creator>
            <pubDate>Tue, 07 Jun 2016 06:07:42 GMT</pubDate>
            <atom:updated>2016-10-30T08:47:06.678Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TDiE4_ANWjtekDNZmisj-g.png" /><figcaption><strong>Try it now — </strong><a href="http://searchstone.io/"><strong>searchstone.io</strong></a></figcaption></figure><p><strong><em>tl;dr: </em></strong><em>I’m releasing a search engine for </em><a href="http://us.battle.net/hearthstone/en/"><em>Hearthstone</em></a><em> called </em><strong><em>Searchstone</em></strong><em>. It’s </em><a href="http://searchstone.io/"><em>live right now</em></a><em> and it’s </em><a href="https://github.com/shipow/searchstone"><em>open source</em></a><em>. Please enjoy responsibly.</em></p><p><a href="http://us.battle.net/hearthstone/en/">Hearthstone</a>’s popularity is stunning. Is it the worthy heir of chess? As of April there were <a href="http://www.statista.com/statistics/323239/number-gamers-hearthstone-heroes-warcraft-worldwide/">50 million players</a> worldwide… it really makes you wonder ;) I must confess that I find <a href="https://www.twitch.tv/directory/game/Hearthstone%3A%20Heroes%20of%20Warcraft">streamed Hearthstone matches</a> more entertaining than soccer. For me HS is a truly interactive global sport.</p><p>One big reason why I love Hearthstone is the community. On GitHub you can find projects dedicated to theorycrafting or creating a competitive artificial intelligence. Today it’s my turn to give something back.</p><p><em>::sounds trumpet::</em></p><p><strong>I’m releasing an open source search engine for Hearthstone. </strong>(<a href="https://github.com/shipow/searchstone">code</a>)</p><p>It’s called <a href="http://searchstone.io/"><strong>Searchstone</strong></a> and you can go try it right now.<strong> </strong>Or you can stay here and learn about how I built it in GLORIOUS detail. I squeezed it into just NINE parts:</p><p>1. <strong>Dissecting the current search</strong><br>2. <strong>Getting the data</strong><br>3. <strong>Pushing the data to </strong><a href="https://www.algolia.com/?utm_medium=link&amp;utm_source=medium.com&amp;utm_campaign=searchstone"><strong>Algolia</strong></a><br>4. <strong>Building the cards</strong><br>5. <strong>Building the user interface</strong><br>6. <strong>Crafting the look and feel</strong><br>7. <strong>Improving relevance</strong><br>8. <strong>Gathering feedback</strong><br>9. <strong>Feature showcase</strong></p><h3>PREAMBLE — What is Algolia?</h3><p>It’s a developer-friendly, UX-obsessed search-engine-as-a-service. It’s part of the alchemy behind Searchstone. <a href="https://www.algolia.com/?utm_medium=link&amp;utm_source=medium.com&amp;utm_campaign=searchstone">Algolia</a> runs in the cloud and comes with front-end JavaScript tools so you can build browser-based search experiences with just a few lines of code. Algolia’s <a href="https://www.algolia.com/users/sign_up/hacker?utm_medium=link&amp;utm_source=medium.com&amp;utm_campaign=searchstone">hacker plan</a> is free and you can run <strong>100,000</strong> queries a month before hitting any limits. <em>Disclaimer: I work at Algolia so of course I’m going to tell you how great it is. To get more opinions you can </em><a href="https://twitter.com/search?q=algolia&amp;src=typd"><em>search for Algolia on twitter</em></a><em> or on </em><a href="https://hn.algolia.com/?query=algolia&amp;sort=byPopularity&amp;prefix=false&amp;page=0&amp;dateRange=all&amp;type=story&amp;experimental"><em>Hacker News</em></a><em>.</em></p><h3>1. Dissecting the current search</h3><p>If you already play Hearthstone then you’re familiar with the following screen. This is the single entry point for building a deck. It’s also the <strong>only official place</strong> to browse through all the cards and research their properties.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z25KyD5IIjaWirhxoO65ng.png" /><figcaption>Hearthstone collection view</figcaption></figure><p>This presents a handful of difficulties.</p><ul><li>My game is installed in French but deck sites like <a href="http://www.hearthpwn.com/">Hearthpwn</a> and streamed games from <a href="https://www.twitch.tv/">Twitch</a> are in English. <strong>The search doesn’t know the names of cards in English if the game is installed in another language.</strong></li><li><strong>I have to press enter each time to launch the search. </strong>The query field doesn’t allow editing a query. Instead, a select-all is applied and overwrites everything. When clicking the cross the search field is erased and loses focus. On a tablet this is really a pain.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/225/1*EyOJeO3va4Hw3Y0COXYEOQ.gif" /></figure><ul><li><strong>The search field prevents clicking on any card when it is selected. </strong>And the left click on a card brings up an error message above the search field!</li><li><strong>The advanced syntax a.k.a. ‘tag system’ (attack 1–3) isn’t explained anywhere</strong> except in <a href="http://us.battle.net/hearthstone/en/blog/20056284/search-your-hearthstone-collection-like-never-before-3-17-2016">a single blog post</a> and it is not very easy to use. I admit this is not a killer feature but it can be handy sometimes.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/327/1*Om5w7viHz_VVDKvyoSsppQ.gif" /><figcaption>click to open, scroll, click to close</figcaption></figure><ul><li><strong>Information on the origin of cards is not obvious and the filters are hidden.</strong> Why? Does God hate me?</li><li><strong>There is no tolerance for typos and spelling mistakes</strong>. A query like <em>Hooger</em> will return nothing. For people not very comfortable with troll naming (apparently these people still exist) that’s pretty lame.</li></ul><h3>2. Getting the data</h3><p>I like JSON because it’s less horrible than CSV or XML. Finding some JSON that contains the cards and their attributes would be nice. Fortunately there is this awesome project called <a href="http://hearthstonejson.com/">hearthstonejson</a>. They have a clean extraction of cards and attributes in all languages.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/739/1*0OmdfHlrp0mGCT2aneaF9w.png" /><figcaption>hearthstonejson is a great project. We’ll use only the collectible cards.</figcaption></figure><p>The JSON could be used without any modification in Algolia but there will still be some optimization to create the perfect search experience. The project’s source code can extract the data to the desired format from an existing game install… but it’s written in Python. So I’ll need to post-process the JSON with a Node.js script.</p><p>This way I can easily work with the data throughout the entire application-building process and use regular expressions to remove markers on certain strings. I can also add loops to split the records by language:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QWS52f4Ro7NjSvoIZ9dxAA.png" /></figure><h3>3. Pushing the data</h3><p>Here are a few illustrations from the <a href="https://algolia.com/doc">Algolia docs</a> that help explain some basic API concepts &amp; vocabulary:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/621/1*xuX32Q7ylrecEItivkVk-A.png" /></figure><p>The first step of configuring an Algolia search is deciding which <a href="https://www.algolia.com/doc/indexing/formatting-your-data">attributes to index</a>. These are the attributes that will be searched when a user performs a query. To start configuring these attributes, I can simply drag &amp; drop a JSON file into my Algolia dashboard. The data will immediately be indexed and displayed on the <em>Browse Tab</em><strong>. This tab allows us to test and adjust the configuration of the index.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NNXGDV7n-f06PCfOqZzN_Q.png" /><figcaption>Uploading JSON via the Algolia dashboard</figcaption></figure><h4>Attributes to index</h4><p>The Algolia engine lets the user search inside of any configured attribute of the record. The <em>name</em> of the HS card is the most important attribute to match, but we also want to match against <em>description</em>, <em>race</em>, etc.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/829/1*d8-F8v-60ME1jU4en3yD3g.png" /><figcaption>Point, click and drag to configure the attributes to index</figcaption></figure><p>There’s one very important thing to keep in mind here. We want to make sure that the content of the attribute matches what is displayed in the results. <strong>This way users can quickly understand <em>why</em> records have been selected (and why some records are ranked <em>before</em> others).</strong> Most of the time this information is textual in nature and clearly indicated with visual highlighting. But not always — a color attribute might be an exception to the textual rule.</p><h4>Attributes for faceting</h4><p><a href="https://www.algolia.com/doc/search/filtering-faceting">Faceting</a> is a fancy term for searching within categories. Configuring an attribute to be faceted on is like creating a filter to refine search results.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/259/1*e1Oy2qK-1joDvkYWlpp7Mg.png" /><figcaption>A facet on the “set” attribute generates this menu</figcaption></figure><p>The list of filters (a.k.a. facets) is dynamic. The interface will display only filters available based on the results of the search, the count for each value can also be displayed.</p><p>In our case, the structure is quite simple and we can simply add all the attributes that have numerical values (<em>mana, attack, health</em>) as well as the ‘tags’ of cards (<em>language, set, class, type</em>).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/859/1*PDgAVnEVLb7B-MyM-vlO7Q.png" /><figcaption>Each faceted attribute gives users more options to refine their search</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UyW-vGPQiQytZiqhSvyykw.png" /><figcaption>The Algolia dashboard allows me to quickly experiment with the settings.</figcaption></figure><p><em>Et voilà!</em> Now we have an index with the right attributes and facets configured. There’s just one more challenge before we can start designing the real UI.</p><h3>4. Building the cards</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/743/1*U_CLufVfApmLEdmhNFKV0g.png" /></figure><p>I want to use Algolia’s <a href="https://www.algolia.com/doc/search/autocomplete-textarea#highlighting-variants"><strong>highlighting feature</strong></a> to accent the matching letters of the search <strong>on the cards</strong>. For example, a result for the query ‘<strong>Demon’</strong><em> </em>would be accented and displayed as ‘<strong>Demon</strong>Wrath’. This is tricky with the card’s name because it is curved and curved differently according to the card type (<em>minions, spell or weapon</em>). One way to accomplish this is by using HTML markup to describe a path — not the simplest thing.</p><p>Thankfully, there is a HearthSim community project called <a href="https://github.com/HearthSim/sunwell"><strong>Sunwell.js</strong></a><strong>. </strong>The library generates cards in the browser, curved text and all. Sunwell requires the whole of Hearthstone’s artwork in JPG format. Let me show you how I extracted it.</p><h4>Extracting the artwork</h4><p>If you have a PC at your disposal, good for you! I currently do not. You need one because (at the time of writing) the primary Python library used to extract images on Mac OSX has malfunctioning alpha layers. My workaround was to mount a Windows VM and make/execute <a href="https://github.com/andburn/hs-art-extractor">this extractor</a> (another HearthSim community project).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/771/1*3NTRapSqSRIC_irqMbfTiQ.png" /><figcaption>Extracting the .png from the Unity data package (here with the Python project and a few buggy alpha channels)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*85FcIs4X3fz5X60ATzPUZg.png" /><figcaption>Installing Visual Studio on a Windows 10 VM: Out of my comfort zone /o\</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*ILCaFpAmSKEU-VkDxeG3sQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*p3MGnOOQWZmhl--6V1DVqw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*HMPDrh3PkvPyTlIW_yuzWg.png" /><figcaption>Worth it though — the HD artwork is ridiculously awesome. Thanks artists!</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/444/1*dKl3xNZi_ZEWZcWMCRrweQ.png" /><figcaption>Sunwell will re-assemble these assets</figcaption></figure><p>Sunwell will generate the card through the information transmitted by the Algolia record. But for <em>mana, name, description</em> and other information that does not need to appear we can just send an empty strings. (Later I removed this step out of the Sunwell library to accelerate the rendering of cards.)</p><h4>Image Hosting</h4><p>Once created, the images are hosted on <a href="http://cloudinary.com/">Cloudinary</a>, an API that allows your assets to be served on the fly with configurable size and quality. The images also get cached on a CDN for faster download speeds.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/344/1*VJwl_amCGMbGw4hLrzyw0g.png" /></figure><p>Each piece of artwork is uploaded as a full-resolution PNG, but the Searchstone UI requests a compressed 300x300 JPG version to display to users. This optimization is important because we want to display many cards simultaneously on the results page as quickly as possible.</p><h4>Rendering</h4><p>We first generate an empty card to use as a background, and then place information on top of it with CSS absolute positioning. Additional CSS classes are used to change the size of the text so that it always remains within the limits (there are very long strings, especially in Russian and Thai).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/314/1*yhGVJkj5ZQ-6RvzisatlFQ.png" /><figcaption>Empty card background with the HTML template on top</figcaption></figure><p>To reproduce the curvature of the text, I create a <em>&lt;svg&gt;</em> element and add a tag <em>&lt;text path&gt;</em> that draws the text along one of the three paths (remember, these depend on the type of card). To produce the highlighting I have to change the <em>&lt;em&gt;</em> to a real SVG text tag <em>&lt;tspan&gt;</em>. It’s possible thanks to the <em>transformData</em> method of the widget.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/785/1*TcPkeCj1lcDkpKttHFf3zg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/267/1*7qbrg6UWMwYdnwSe1DI2qA.png" /><figcaption>Path for Minion Cards</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/265/1*oKUfKZmyWsG2RLPmke1EDg.png" /><figcaption>Path for Spells</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/890/1*Tv66lvnzT-ApAwi6FeAQyw.png" /><figcaption>Three SVG path definitions</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IwSAMpjZnWigGzbLQMDT0Q.png" /><figcaption>The final HAML template of a card. {{textPath}} specifies the right path for the card type.</figcaption></figure><p><em>Ta-da!</em> Here’s a look at the final card rendering, including highlighting:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/922/1*i7BSO6KSDc91gq8cUW1mUg.gif" /></figure><h3>5. Building the user interface</h3><p>The UI is built with <a href="https://community.algolia.com/instantsearch.js/">instantsearch.js</a>, an open source JavaScript library created and actively maintained by Algolia. It’s based on <a href="https://facebook.github.io/react/">React</a> but you don’t need to know React to use it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/424/1*dLYoIUGshPB8n2SLt7Whgw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/296/1*SSxIQeLJ4k3e4OMBmiOIEw.png" /></figure><p>instantsearch.js generates all of the UI elements that you need to offer a great search. This includes widgets for typing in the query, displaying results and exposing extras like filters and pagination.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Lz4KzIR8iCDeMHsKpHl2XQ.png" /><figcaption>An instantsearch.js documentation sample; the code on the right generates the UI on the left</figcaption></figure><p>As part of the design process I asked HS players — <em>what are the most essential attributes and their priorities?</em> Then I took tips from the in-game navigation and other acclaimed HS search sites.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eEH6x0fbCWRkTRIsz9o-LA.png" /><figcaption>In the very first version images came directly from zamgaming (using the card ID as a URL parameter)</figcaption></figure><p>The mobile version required a bit more thought. I used <a href="https://www.sketchapp.com/">Sketch</a> to mockup various compromises between space optimization, readability, and functionality, as well as the breakpoint transformations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/604/1*S4sDJYB-l5kzESyqwEEPgQ.gif" /><figcaption>Designing the open filters menu. (Sketch pro-tip: Press Alt and Drag and Drop, then Cmd+D)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LAwN_XKFK8UyubZnBhDBQw.png" /><figcaption>UI needed more attention. Sketch is perfect companion for this!</figcaption></figure><h4>JavaScript development environment</h4><p>I use <a href="http://gulpjs.com/">Gulp</a> to run tasks and I am quite attached to it, having used it in many other projects. For bundling and minifying the JavaScript I used <a href="https://webpack.github.io/">webpack</a>. Webpack isn’t one of my top strengths yet but that’s ok, one of the reasons I do projects like Searchstone is to have an excuse to try new stuff :)</p><h3>6. Crafting the look and feel</h3><p>The artwork from the game is nice but I wanted to add a few special touches. For one, I reproduced the existing icons in vector format to make rendering cleaner.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*o-juwxxncI1do_odscEm_A.png" /><figcaption>Sketch makes things so simple ❤</figcaption></figure><h4>Two views: full and compact</h4><p>There are two layouts that the user can switch between at any time. The full, grid-based view shows each card in its traditional shape and at a sufficient size to see the artwork and read the description. When you click on a card it animates and comes into the foreground.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7fhw_6ikqZahtnNyYuCGDw.png" /><figcaption>Full view; the card details could be enriched with some data and stats!!!</figcaption></figure><p>A compact view allows the user to skim information faster and more easily compare attributes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*gwXDbkj40299kJMl8pXk3w.png" /><figcaption>Compact view</figcaption></figure><h3>7. Improving Relevance</h3><p>Relevance is a very important concept when it comes to building search applications. A relevant search returns what the user wants in a predictable, sensible order. The ability to tune relevance at a fine-grained level is one of the things I like best about Algolia. I also like that tuning doesn’t require writing code. It can be done by clicking, dragging and dropping inside your Algolia dashboard.</p><p>I feel confident configuring relevance but then I also know a few people whose skills are <em>diabolically precise</em>. One of these people is my co-worker <a href="https://twitter.com/nicolasbaissas">Nicolas Baissas</a>, or NicoB for short. He works on the Customer Solution Engineering team, or CSE for short. The Algolia CSEs design solutions for the most complicated search scenarios. They also <a href="https://blog.algolia.com/how-algolia-tackled-the-relevance-problem-of-search-engines/">write about relevance</a> on our blog.</p><p>This was one of my favorite moments of the project because I got to learn from NicoB about the intricate world of search performance and relevancy. And I do mean intricate… one of our Skype sessions lasted over two hours :)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/658/1*weH80O-L_MkwTRxKz9QWDg.png" /></figure><h4>Industrial-strength relevancy</h4><p>During our conversation we discussed and configured options related to <a href="https://www.algolia.com/doc/relevance/typo-tolerance">typo tolerance</a>, <a href="https://www.algolia.com/doc/relevance/ranking">ranking strategy</a>, <a href="https://www.algolia.com/doc/relevance/sorting">sorting</a>, <a href="https://www.algolia.com/doc/indexing/formatting-your-data">record size</a>, <a href="https://www.algolia.com/doc/relevance/synonyms">synonyms</a> and optional words. NicoB had already played HS so he was familiar with a lot of the attributes and that helped a lot.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*01a1ipWEVbU_0DU2D6kM7A.png" /><figcaption>Configuring optional words in the Algolia dashboard</figcaption></figure><h3>8. Getting feedback</h3><p>After the first version was done I sent the link around to some friends and players to get some feedback.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*jXEqiCI8FZgrMF-xuh35XQ.png" /><figcaption>Useful feedback about the UI from Dribbble</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/726/1*7z3PsUienZ7TI1An8OQJ2g.png" /><figcaption>Feedback from the battle.net forum</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/725/1*ZE68pbbXRQXV1tTiXt2acw.png" /><figcaption>More battle.net feedback</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EG4gjfNqgPyFikzuNRewkw.png" /><figcaption>Some users couldn’t see the cards on Firefox :/</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/618/1*qVPclel0fPNlKMA8TyBLpg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/623/1*hvdn2nTJpzKTtF6a61_tyQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/599/1*LsWovjpC9z5ogPDffKeZ5Q.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*M77hDW_Bh-DEJwRH2fOUcQ.png" /><figcaption>Mobile / Tablet / Desktop</figcaption></figure><h3>9. Feature showcase</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fH9TPod6XULqt-7-6x5p4w.gif" /><figcaption>Search-as-you-type experience</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/237/1*YpznMg5gkZAW56YzZNatYA.gif" /><figcaption>Filter by Set and Wild/Standard format</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/305/1*mMYO_P2zWiqPhG5RW-jDUQ.gif" /><figcaption>Support for multiple languages</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/852/1*VUa0EEQ3Z2Qr0tYHAWhrVA.png" /><figcaption>Typo-tolerance for spelling mistakes</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*tmxfwNBCmXLaRvDxtP9jMw.gif" /><figcaption>Golden Cards Animations</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/462/1*Jc-YJU009lMUJH24Nm1M2Q.gif" /><figcaption>Attack/Health range sliders</figcaption></figure><h3>EPILOGUE — Feedback, please!</h3><p>Now that the current version is up and running I’d love to get more feedback from players of all levels. If you know a few players that would benefit, <strong>could you </strong><a href="http://searchstone.io/"><strong>share Searchstone</strong></a><strong> with them? :)</strong></p><p>The project will hopefully join the HearthSim community soon, and may benefits from the super interesting stats in <a href="https://github.com/HearthSim/Hearthstone-Deck-Tracker">their tracker</a>. I think this could lead to really cool stuff, like:</p><ul><li>Show golden animated cards.</li><li>Show proposed cards from the ‘discover’ mechanic in the detail view.</li><li>Improve the <em>No Results</em> page.</li><li><strong>Share your ideas</strong> as Medium comments or <a href="https://github.com/Shipow/searchstone/issues">GitHub issues</a>.</li></ul><h3>Many Thanks</h3><ul><li>All of the <strong>Algolia team</strong> who helped make this possible. Especially <strong>Nicolas Baissas</strong> and <strong>Tim Carry.</strong></li><li><strong>Jerome Leclanche</strong> for guiding me around the Hearthstone community and for the hearthstoneJSON project.</li><li><strong>Christian Engel</strong> for creating Sunwell, and helping.</li><li><strong>Andrew Burnett</strong> for the art extractor and the Windows 10 free tour.</li><li><strong>Jean-Francois Fraisse </strong>for his great insights on the mobile version.</li><li><strong>Gabe, Josh </strong>&amp;<strong> Lucas</strong> for proofreading this article!</li><li><strong>Blizzard</strong>, because you are f* good</li><li><strong>All illustrators involved in the creation of these cards.</strong> When I see your work I feel a bit nostalgic that I left that path. Please keep it up and continue to blow our minds :)</li><li><strong>Dribbble </strong>and<strong> Twitter users</strong> for your support and (seriously) useful feedback.</li></ul><h3>APPENDIX: HS Community Showcase</h3><p>These are some screenshots and links to parts of the HS community I’ve benefitted from. Maybe you can too :)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2qCH7vX01A26zphPvXoctQ.png" /><figcaption><a href="http://www.hearthpwn.com/cards?display=3">http://www.hearthpwn.com/cards?display=3</a></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AELrKsGeGJU2_lOES7HDZA.png" /><figcaption><a href="http://www.hearthhead.com/cards">http://www.hearthhead.com/cards</a></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/984/1*8RKerdHqh0kZIGArXrPKCg.png" /><figcaption><a href="http://tempostrom.com/">http://tempostrom.com</a></figcaption></figure><ul><li><a href="http://arenadrafts.com/">Arena Drafts</a></li><li><a href="http://hearthstone.gamepedia.com/Hearthstone_Wiki">Hearthstone Wiki</a></li><li><a href="http://hearthstoneplayers.com/">Hearthstone Players - Hearthstone Decks and Guides</a></li><li><a href="http://www.heartharena.com/tierlist">HearthArena&#39;s Hearthstone Arena Tier List</a></li><li><a href="http://www.hearthstonetopdecks.com/">Hearthstone Top Decks - The Best Decks on the Net!</a></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/831/1*bDE4gNy-kLJKC8mctAP2cg.png" /><figcaption><a href="http://web.archive.org/web/20000815065659/http://www.battle.net/scc/protoss/">http://web.archive.org/web/20000815065659/http://www.battle.net/scc/protoss/</a></figcaption></figure><blockquote>Dear Blizzard, you made me who I am today. I learned HTML and CSS from the original Starcraft website. Yes, the first version of the game back in 1999. Special thanks to web browsers for having that cool ‘View Source’ feature ;p</blockquote><p><em>— by </em><a href="https://twitter.com/Kevin_Granger"><em>Kevin Granger</em></a><em>, Devsigner @</em><a href="https://medium.com/u/a531947eec97"><em>Algolia</em></a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/129/1*4bdpY7UwJhNo4Jv5C1FCtQ.png" /><figcaption><a href="https://twitter.com/Kevin_Granger">Let’s be friends!</a></figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c21b3fa4223c" width="1" height="1" alt=""><hr><p><a href="https://medium.com/algolia-stories/a-painstakingly-crafted-search-for-hearthstone-c21b3fa4223c">A painstakingly crafted search for Hearthstone</a> was originally published in <a href="https://medium.com/algolia-stories">Algolia Stories</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>