<?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 Cedric Bouletkessler on Medium]]></title>
        <description><![CDATA[Stories by Cedric Bouletkessler on Medium]]></description>
        <link>https://medium.com/@cedric.bouletkessler?source=rss-60efb25726a2------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*yfj9vIsV-xbefD8F</url>
            <title>Stories by Cedric Bouletkessler on Medium</title>
            <link>https://medium.com/@cedric.bouletkessler?source=rss-60efb25726a2------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 18:44:13 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@cedric.bouletkessler/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[How We Transformed Doctrine’s Entire UI in a Month — Without Any Downtime]]></title>
            <link>https://medium.com/doctrine/how-we-transformed-doctrines-entire-ui-in-a-month-without-any-downtime-c3dd02c4df8c?source=rss-60efb25726a2------2</link>
            <guid isPermaLink="false">https://medium.com/p/c3dd02c4df8c</guid>
            <category><![CDATA[react]]></category>
            <category><![CDATA[migration]]></category>
            <category><![CDATA[design-systems]]></category>
            <category><![CDATA[redesign]]></category>
            <category><![CDATA[engineering]]></category>
            <dc:creator><![CDATA[Cedric Bouletkessler]]></dc:creator>
            <pubDate>Wed, 05 Nov 2025 13:27:20 GMT</pubDate>
            <atom:updated>2025-11-06T15:11:18.827Z</atom:updated>
            <content:encoded><![CDATA[<h3>How We Transformed Doctrine’s Entire UI in a Month — Without Any Downtime</h3><h3>The Day We Decided to Redesign Everything</h3><p>Rebuilding a live product’s design, while it’s in full use, feels a little like remodeling a plane while it is mid-air — it’s risky, but feels great when it works!</p><p>And that’s exactly what we wanted to do.</p><p>At Doctrine, we have spent the past eight years building the first AI platform for legal professionals, first in France and now Europe. The product has been continually evolving — from being a legal search engine into a productivity suite. But because of the new tools and workflows, our user interface was starting to show its age. Some pages looked years apart from others, and we found ourselves facing increasing difficulty incorporating those new tools and workflows.</p><p>So we made the decision to revamp the visual identity of both doctrine.fr and doctrine.it, in just a month, with three developers, without any downtime, and whilst fifteen other engineers simultaneously kept shipping features every day (at Doctrine we do CI/CD, meaning multiple small releases are sent to production every day).</p><p>I had been part of a redesign process a few years before for a much smaller platform, and chaos would be an understatement — long-lived branches, hell on rebase, and a partially frozen product for weeks. This time we promised to do better.</p><p>We wanted speed, to ship with confidence, and keep the product stable in production until ready to flip a switch to show the brand new interface to the world (or at least Europe).</p><p>Just a simple process, a solid design system, and one clever React pattern.</p><h3>The Real Foundation: a Well-Built Design System</h3><p>The reality is, we started working on foundations that allowed this project long before the redesign idea came up.</p><p>We spent a year and a half working on our own internal design system, developing components, CSS conventions, documentation, deduping similar components, extracting business logic from presentation, and so on.</p><p>It was not sexy work, but it resulted in a solid foundation for our new endeavour and the investment paid off.</p><p>When the new visual identity emerged, the building blocks were there.</p><p>We were able to refresh the entire visual identity of both sites, in large part because of this groundwork.</p><h3>The Plan: a Higher-Order Component and a Feature Flag</h3><p>The true unsung heroes of this story were a fairly small React pattern — a <a href="https://legacy.reactjs.org/docs/higher-order-components.html">Higher-Order Component</a> (HOC) — and a <a href="https://martinfowler.com/articles/feature-toggles.html">feature flag</a>.</p><p>Each element needing refreshing was updated to support the two display formats. If the change was easy enough, we could change only the CSS class applied ; when the look and feel was too different we created the full v2 component ; or if it simply didn’t exist before or after, a specific version could return null. The HOC would be responsible for determining which one to render based on the feature flag for the user being logged in.</p><p>The HOC definition:</p><pre>export function ApplyNewGlobalNavHoc&lt;T extends object, U extends object&gt;(<br>  OldComponent: FC&lt;T&gt;,<br>  NewComponent: FC&lt;U&gt;<br>): FC&lt;Partial&lt;T &amp; U&gt;&gt; {<br>  const WrappedComponent: FC = (props) =&gt; {<br>    const { featureFlags } = useUser();<br>    return featureFlags.includes(&#39;doctrine-v2&#39;) ? (<br>      &lt;NewComponent {...(props as U)} /&gt;<br>    ) : (<br>      &lt;OldComponent {...(props as T)} /&gt;<br>    );<br>  };<br><br>  return WrappedComponent;<br>}</pre><p>With the HOC, we could merge PRs under main on a regular basis — even if the redesign wasn’t fully baked. Users who were beta testers with the feature flag enabled saw the partial new interface, and everyone else saw the old interface in their app.</p><h3>No Merge Hell, No Stress.</h3><p>If you’ve ever worked on a large front-end revamp, you know the pain: the longer a branch lives, the more conflicts and rebases you’ll face, and more chance you’ll lose or break something.</p><p>Our HOC + feature flag combo solved that pain.</p><p>No more endless rebases or hanging branches.</p><p>We could ship continuously, review code as usual, and keep production clean and stable.</p><p>Better yet, other developers could update v1 &amp; v2 when they shipped code as they saw the two variants on the main branch.</p><p>We even did a gradual rolled out some pages before the redesign was complete, when the cohabitation was not possible.</p><p>The transition was smooth, controlled, and it was invisible to most users!</p><h3>One Easy but Important Convention: the good old V2</h3><p>To keep things clear, we established a naming convention: every component or CSS class from the new UI would have a “V2” suffix at the end of the name.</p><ul><li>Button → ButtonV2</li><li>.header → .headerV2</li></ul><p>That simple convention saved us a lot of mistakes.</p><p>It made the coexistence of both versions explicit, helped everyone when they edited code to know which version they were editing, and helped change both when appropriate.</p><p>Usage example:</p><pre>import styles from &#39;./FullPageCard.module.scss&#39;;<br><br>const FullPageCardV1: FC&lt;PropsWithChildren&gt; = ({ children }) =&gt; {<br>  return (<br>    &lt;PageContainer classNames={styles.card}&gt;<br>      &lt;Page classNames={styles.page}&gt;<br>        &lt;Card className={styles.paper}&gt;{children}&lt;/Card&gt;<br>      &lt;/Page&gt;<br>    &lt;/PageContainer&gt;<br>  );<br>};<br><br>const FullPageCardV2: FC&lt;PropsWithChildren&gt; = ({ children }) =&gt; {<br>  return (<br>    &lt;Page classNames={styles.pageV2}&gt;<br>      &lt;Card className={styles.paper}&gt;{children}&lt;/Card&gt;<br>    &lt;/Page&gt;<br>  );<br>};<br><br>export const FullPageCard = ApplyNewGlobalNavHoc(<br>  FullPageCardV1,<br>  FullPageCardV2<br>);</pre><pre>.page {<br>  justify-content: center;<br>  align-items: center;<br>  flex-direction: column;<br>  justify-content: center;<br>}<br><br>.pageV2 {<br>  justify-content: center;<br>  flex-direction: column;<br>  align-items: center;<br>  justify-content: center;<br><br>  margin-top: global.$spacing-unit * 4;<br>}</pre><h3>Feature Flags: Secure Release, Easy Testing</h3><p>The feature flag was not only an assurance for code safety, but also enabled a staged rollout.</p><p>We first exposed the new UI bit to internal testers, then to a few selected beta users, then to everyone.</p><p>This “silent release” gave us a chance to refine details, fix some small layout issues, and observe performance before we exposed it globally.</p><h3>The AI-Powered Cleanup</h3><p>After the new UI had been out there for everyone to use, there was just one final step: to remove the v1 stuff.</p><p>To be frank, nobody likes that.</p><p>This is where AI tools like Copilot and Cursor came in.</p><p>We used them to automate repetitive tasks around cleanup: identifying v1 only components and classes, removing them and then dropping v2 namings.</p><p>Admittedly, it was not the most fun work; however, it was fast and mostly painless since we could use AI tools to assist us.</p><p>We tried both Copilot and Cursor to be able to compare them afterward.</p><p>The prompt more or less was as follows:</p><pre>For this component:<br>1. If a SCSS module is associated: compare v1 and v2 of the component, then remove styles from the associated SCSS module that are only used for v1. Do not touch v2 styles nor merge styles together.<br>2. Keep only v2 of the component and rename it without the suffix, then export it as &quot;export const.&quot;<br>3. If the component uses a v2 style class, remove the suffix from the component in the associated SCSS module.<br>Do not add new comments to the code.</pre><p>Cursor was able to edit everything in a folder at once, but it made a lot of mistakes, added random things, and the diff that resulted from that confusion often made no sense.</p><p>Copilot only edited one component at a time, but made less mistakes, and the diff opt-out was clearer and easier to read most of the time.</p><h3>One Month In: Mission Accomplished</h3><p>In a span of four weeks we had:</p><ul><li>Migrated the entire UI for two live sites.</li><li>No major regressions in production.</li><li>Cleared out the legacy components.</li><li>Maintained a smooth and conflict-free merge flow.</li></ul><p>No late-night deploys, no code freezes, no Git nightmares.</p><p>Just a thoughtful plan, smart tools, and a coordinated team.</p><h3>What We Learned</h3><p>What we learned about the redesign is quite simple:</p><p>A smooth migration isn’t about speed — it’s about preparation (your design system), strategy (your feature-flagged HOC), and discipline (your conventions and cleanup).</p><p>And with just a little help from AI, you can do it without pain.</p><h3>Bonus</h3><p>The mandatory before and after</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4EzusH70AypEQWvhFFm1SA.png" /><figcaption>(before)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2Rx7RcvbU5bYgxPMp8hyUg.png" /><figcaption>(after)</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c3dd02c4df8c" width="1" height="1" alt=""><hr><p><a href="https://medium.com/doctrine/how-we-transformed-doctrines-entire-ui-in-a-month-without-any-downtime-c3dd02c4df8c">How We Transformed Doctrine’s Entire UI in a Month — Without Any Downtime</a> was originally published in <a href="https://medium.com/doctrine">Inside Doctrine</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>