<?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 Hugorus on Medium]]></title>
        <description><![CDATA[Stories by Hugorus on Medium]]></description>
        <link>https://medium.com/@hugorus87?source=rss-95a07a694494------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*6BBtw-2L-Ai_Tpg3</url>
            <title>Stories by Hugorus on Medium</title>
            <link>https://medium.com/@hugorus87?source=rss-95a07a694494------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 23:58:07 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@hugorus87/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 to Add Email Authentication to an Expo App Using Supabase]]></title>
            <link>https://medium.com/@hugorus87/how-to-add-email-authentication-to-an-expo-app-using-supabase-88f1f74668d7?source=rss-95a07a694494------2</link>
            <guid isPermaLink="false">https://medium.com/p/88f1f74668d7</guid>
            <category><![CDATA[tinbase]]></category>
            <category><![CDATA[supabase]]></category>
            <category><![CDATA[react-native]]></category>
            <category><![CDATA[typescript]]></category>
            <category><![CDATA[mobile-development]]></category>
            <dc:creator><![CDATA[Hugorus]]></dc:creator>
            <pubDate>Mon, 16 Nov 2026 14:00:11 GMT</pubDate>
            <atom:updated>2026-11-16T14:00:11.277Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Supabase</strong> is an open-source backend that provides a Postgres database, file storage and user authentication. In this article, we will add email and password sign-up, sign-in and sign-out to a React Native app built with <strong>Expo</strong>, and keep users logged in after the app restarts.</p><h3>Prerequisites</h3><ul><li>Basic knowledge of React Native and hooks</li><li>Node.js installed</li><li>A free Supabase project (from supabase.com)</li></ul><h3>Step 1: Create the Project and Install Libraries</h3><pre>npx create-expo-app@latest AuthDemo<br>cd AuthDemo<br>npx expo install @supabase/supabase-js @react-native-async-storage/async-storage react-native-url-polyfill</pre><ul><li><strong>@supabase/supabase-js:</strong> the Supabase client.</li><li><strong>async-storage:</strong> saves the login session on the device.</li><li><strong>url-polyfill:</strong> adds URL support the client needs.</li></ul><h3>Step 2: Create the Supabase Client</h3><p>In your Supabase dashboard, open <strong>Project Settings → API</strong> and copy the project URL and the anon (public) key. Create supabase.js:</p><pre>// supabase.js<br>import &#39;react-native-url-polyfill/auto&#39;;<br>import AsyncStorage from &#39;@react-native-async-storage/async-storage&#39;;<br>import { createClient } from &#39;@supabase/supabase-js&#39;;<br>import { AppState } from &#39;react-native&#39;;</pre><pre>export const supabase = createClient(&#39;YOUR_PROJECT_URL&#39;, &#39;YOUR_ANON_KEY&#39;, {<br>  auth: {<br>    storage: AsyncStorage,     // keep the session after restarts<br>    autoRefreshToken: true,<br>    persistSession: true,<br>    detectSessionInUrl: false, // there is no browser URL in a mobile app<br>  },<br>});</pre><pre>// Refresh tokens only while the app is open<br>AppState.addEventListener(&#39;change&#39;, (state) =&gt; {<br>  if (state === &#39;active&#39;) supabase.auth.startAutoRefresh();<br>  else supabase.auth.stopAutoRefresh();<br>});</pre><p><strong>Note:</strong> the anon key is safe to include in the app only if <strong>Row Level Security</strong> is enabled on your tables. Never put the service role key in an app.</p><h3>Step 3: Build the Login Screen and Session Handling</h3><p>Replace the contents of App.js:</p><pre>// App.js<br>import { useEffect, useState } from &#39;react&#39;;<br>import { View, Text, TextInput, Button, Alert } from &#39;react-native&#39;;<br>import { supabase } from &#39;./supabase&#39;;</pre><pre>export default function App() {<br>  const [session, setSession] = useState(null);<br>  const [email, setEmail] = useState(&#39;&#39;);<br>  const [password, setPassword] = useState(&#39;&#39;);</pre><pre>  useEffect(() =&gt; {<br>    // Load a saved session, then listen for changes<br>    supabase.auth.getSession().then(({ data }) =&gt; setSession(data.session));<br>    const { data } = supabase.auth.onAuthStateChange((_event, s) =&gt; setSession(s));<br>    return () =&gt; data.subscription.unsubscribe();<br>  }, []);</pre><pre>  async function signUp() {<br>    const { error } = await supabase.auth.signUp({ email, password });<br>    if (error) Alert.alert(error.message);<br>    else Alert.alert(&#39;Check your email to confirm your account&#39;);<br>  }</pre><pre>  async function signIn() {<br>    const { error } = await supabase.auth.signInWithPassword({ email, password });<br>    if (error) Alert.alert(error.message);<br>  }</pre><pre>  if (session) {<br>    return (<br>      &lt;View style={{ flex: 1, justifyContent: &#39;center&#39;, padding: 24, gap: 12 }}&gt;<br>        &lt;Text&gt;Logged in as {session.user.email}&lt;/Text&gt;<br>        &lt;Button title=&quot;Sign out&quot; onPress={() =&gt; supabase.auth.signOut()} /&gt;<br>      &lt;/View&gt;<br>    );<br>  }</pre><pre>  return (<br>    &lt;View style={{ flex: 1, justifyContent: &#39;center&#39;, padding: 24, gap: 12 }}&gt;<br>      &lt;TextInput placeholder=&quot;Email&quot; value={email} onChangeText={setEmail}<br>        autoCapitalize=&quot;none&quot; keyboardType=&quot;email-address&quot; /&gt;<br>      &lt;TextInput placeholder=&quot;Password&quot; value={password} onChangeText={setPassword}<br>        secureTextEntry /&gt;<br>      &lt;Button title=&quot;Sign in&quot; onPress={signIn} /&gt;<br>      &lt;Button title=&quot;Sign up&quot; onPress={signUp} /&gt;<br>    &lt;/View&gt;<br>  );<br>}</pre><p><strong>Explanation:</strong></p><ul><li>getSession loads the saved session when the app starts, so users stay logged in.</li><li>onAuthStateChange updates the screen whenever the user signs in or out.</li><li>signUp creates an account. By default, Supabase sends a confirmation email first.</li><li>signInWithPassword and signOut log the user in and out.</li></ul><p>Run the app with npx expo start.</p><h3>Output</h3><p>The app shows email and password fields. After signing up and confirming the email, signing in shows “Logged in as” with the user’s email. Closing and reopening the app keeps the user logged in, and <strong>Sign out</strong> returns to the login form.</p><h3>Conclusion</h3><p>With Supabase, an Expo app gets secure email authentication in three steps: install the client, configure it to store sessions with AsyncStorage, and use signUp, signInWithPassword and onAuthStateChange to manage users. Next, enable Row Level Security so each user can only access their own data.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=88f1f74668d7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Mobile App Development Platform (MADP): The 2026 Guide]]></title>
            <link>https://medium.com/@hugorus87/mobile-app-development-platform-madp-the-2026-guide-896668b97483?source=rss-95a07a694494------2</link>
            <guid isPermaLink="false">https://medium.com/p/896668b97483</guid>
            <category><![CDATA[build-mobile-apps]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[mobile-app-development]]></category>
            <category><![CDATA[mobile-apps]]></category>
            <category><![CDATA[madp]]></category>
            <dc:creator><![CDATA[Hugorus]]></dc:creator>
            <pubDate>Mon, 02 Nov 2026 06:01:19 GMT</pubDate>
            <atom:updated>2026-11-02T06:01:19.331Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Originally published at </em><a href="https://www.rapidnative.com/blogs/mobile-app-development-platform?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=madp-guide&amp;utm_content=originally-published"><em>rapidnative.com</em></a><em> by Sanket Sahu.</em></p><p><em>What a mobile app development platform is, how the category went from MEAP to MADP to low-code to AI, and the eight questions that separate a good platform from an expensive detour.</em></p><p>If you have ever sat through an enterprise software review, you have heard someone ask which mobile app development platform the company should standardize on. The question sounds simple. The answers rarely are, because the category has been renamed four times in fifteen years, and the tools sold under the label range from visual workflow builders to raw frameworks to, now, AI that writes the whole app from a paragraph of text.</p><p>This guide untangles it. It covers what a mobile app development platform (MADP) actually is, where the term came from, the five kinds of platforms you will meet in 2026, and the questions that tell a good one from an expensive detour. We build <a href="https://www.rapidnative.com/?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=madp-guide&amp;utm_content=intro">RapidNative</a>, an AI mobile app builder, so we have a point of view, and we will be clear about where it fits and where it doesn’t.</p><h3>What is a mobile app development platform?</h3><p>A mobile app development platform (MADP) is a set of tools and services for designing, building, testing, publishing and maintaining mobile apps for iOS and Android from one place. A typical MADP combines a front-end builder or framework, backend services, integrations with existing systems, security controls, and a way to distribute and update apps once they are in people’s hands.</p><p>The key word is <em>platform</em>. A framework like React Native or Flutter is one ingredient. A platform wraps the ingredients into a lifecycle: you start a project, connect data, give it to testers, ship it to the App Store and Google Play (or to employees’ managed devices), and keep it updated, without assembling a dozen separate products yourself.</p><h3>From MEAP to MADP to MXDP to LCAP: a short history</h3><p>Most of the vocabulary comes from analyst firms, chiefly Gartner, whose Magic Quadrant reports shaped how enterprises shopped for mobile tools. The names changed as the market did.</p><ul><li><strong>2008 to early 2010s, MEAP (mobile enterprise application platform):</strong> middleware and tools to build and run internal apps for employees, often across BlackBerry, Windows Mobile and early iPhone.</li><li><strong>Around 2010, MCAP (mobile consumer application platform):</strong> the customer-facing sibling of MEAP, a short-lived separate category.</li><li><strong>2012 to 2018, MADP (mobile app development platform):</strong> one category for employee and customer apps, as smartphones went mainstream.</li><li><strong>2019 to early 2020s, MXDP (multiexperience development platform):</strong> apps across mobile, web, wearables, chat and voice from one toolset.</li><li><strong>2019 onward, LCAP (enterprise low-code application platform):</strong> visual, model-driven platforms where mobile is one output among many.</li><li><strong>2023 onward, AI app builders:</strong> platforms that generate the app from natural language.</li></ul><p>A few details are worth knowing. Gartner coined MEAP in a 2008 Magic Quadrant, along with a “rule of three”: a MEAP made sense once a company had three or more mobile apps, three or more mobile operating systems, or three or more backend systems to connect. The category was later renamed MADP to reflect that mobile had gone mainstream. In July 2019 Gartner published its first Magic Quadrant for Multiexperience Development Platforms, which superseded the MADP report. Its leaders were mostly low-code vendors: OutSystems, Mendix, Salesforce, Oracle, Progress and Kony.</p><p>The vendors themselves moved too. Kony was acquired by Temenos in 2019, and its non-banking platform was licensed to HCL and relaunched as HCL Volt MX. OutSystems bought Ionic, the team behind Capacitor. Today most of the old MADP leaders are evaluated as enterprise low-code platforms, and “MADP” survives mainly as a search term and an RFP line item. That is useful to know: when someone asks you for a MADP, they usually mean “one supported way to build our mobile apps”, not a specific product category.</p><h3>What a mobile app development platform includes</h3><p>Whatever the label, buyers expect a MADP to cover the full lifecycle of an app. In practice that breaks into six layers.</p><ol><li><strong>Front-end development.</strong> How screens get built: visual drag and drop, a code framework, generated code, or a mix. This decides how the app looks and feels and how fast your team can change it.</li><li><strong>Backend services.</strong> Data storage, sign-in, file storage, push notifications and realtime updates, often called mobile backend as a service (MBaaS). Some platforms bundle their own; others connect to Firebase, Supabase or your existing APIs.</li><li><strong>Integration.</strong> Connectors to the systems an enterprise already runs: identity providers, ERP and CRM, internal REST or GraphQL APIs.</li><li><strong>Security and management.</strong> Single sign-on, role-based access, data encryption, and support for mobile device management (MDM) or mobile application management (MAM) so IT can control company data on phones.</li><li><strong>Testing and distribution.</strong> Previewing on real devices, beta testing, and publishing to the App Store and Google Play, or privately to employees through Apple Business Manager and managed Google Play.</li><li><strong>Lifecycle and updates.</strong> Over-the-air updates, crash reporting, analytics, and a way to keep the app current as iOS and Android change every year.</li></ol><p>No platform is best at all six. The useful question is which layers you want the platform to own, and which you want to keep in your own hands.</p><h3>The five types of mobile app development platforms in 2026</h3><p>The market has split into five approaches. They differ less in features than in two things: what they output, and what you are left holding if you leave.</p><ul><li><strong>Native SDKs</strong> (Xcode with Swift, Android Studio with Kotlin): two separate native codebases. Suits teams with dedicated iOS and Android engineers. The catch: two of everything, from code to bugs to hiring.</li><li><strong>Cross-platform frameworks</strong> (React Native, Flutter, Kotlin Multiplatform): one codebase, native apps. Suits product teams with mobile developers. The catch: you still assemble backend, CI and distribution yourself.</li><li><strong>Enterprise low-code, LCAP</strong> (OutSystems, Mendix, Microsoft Power Apps, HCL Volt MX): visual apps on the vendor’s runtime. Suits large IT teams building many internal apps. The catch: per-user pricing and runtime lock-in.</li><li><strong>No-code builders</strong> (FlutterFlow, Adalo, Glide): apps assembled visually, some with code export. Suits small teams and simple apps. The catch: hand-building every screen, and limits on custom logic.</li><li><strong>AI app builders</strong> (RapidNative for native mobile; Lovable, Bolt and v0 for web): an app generated from a description. Suits teams that want working software in days. The catch: quality depends on the output being real, ownable code.</li></ul><p>Two observations from building in this space.</p><p><strong>First, the framework layer has been commoditized.</strong> In 2012 a MADP’s main job was solving cross-platform: write once, run on iOS, Android and BlackBerry. React Native and Flutter solved that for free. What a platform sells today is everything around the framework: speed, backend, governance, and the last mile into the stores.</p><p><strong>Second, runtime lock-in is the real dividing line.</strong> Many low-code and no-code platforms run your app inside their own engine. That is fine until pricing changes, a feature you need sits on their roadmap, or the company is acquired. Platforms that output standard source code give you an exit. We wrote more about this in our guide to avoiding vendor lock-in on the RapidNative blog.</p><h3>How to evaluate a mobile app development platform: 8 questions</h3><p>Feature matrices all look the same. These questions don’t.</p><h3>1. Is the output a native app or a website in a frame?</h3><p>Many tools that call themselves mobile platforms produce a web app wrapped in a web view. Users notice the scrolling, the keyboard, the lack of native gestures. Apple notices too: App Store Review Guideline 4.2 rejects apps that are “a repackaged website”. Ask for a build on a real phone, not a desktop browser preview. For the longer version of this argument, see our post on native vs hybrid app development.</p><h3>2. Can you leave with working source code?</h3><p>“Export” can mean a zip that still calls the vendor’s runtime. Ask for the exported project, hand it to a developer who has never used the platform, and see whether they can build and run it with standard tools. If they can’t, you are renting, not owning.</p><h3>3. Where does the data live, and can you point it elsewhere?</h3><p>A bundled backend is great for speed. Check you can also connect your own database, identity provider and APIs, and that data can be exported in a standard format.</p><h3>4. How does it handle identity and device management?</h3><p>For internal apps: SSO through your identity provider, role-based access, and compatibility with your MDM or MAM tooling. For customer apps: sign-in methods, account deletion (which both stores now require), and privacy labels.</p><h3>5. How do apps reach users?</h3><p>Publishing is where many projects stall. Certificates, provisioning profiles, store listings, screenshots, privacy policies and review rejections take weeks if nobody on the team has done it. Ask the vendor to walk you through the path from “done” to “live in both stores”, and to private distribution if you need it.</p><h3>6. How are updates shipped?</h3><p>Can you push a fix without a full store review? Frameworks like Expo support over-the-air updates for JavaScript changes within the store rules. Native changes still go through review.</p><h3>7. Who on your team can use it?</h3><p>If only certified specialists can work in the tool, it becomes a bottleneck. The best platforms let a product manager or designer make real progress while engineers stay in control of the codebase.</p><h3>8. What does it cost at 10x the users?</h3><p>Per-user licensing that looks cheap for a 50-person pilot can be the biggest line item at 5,000 employees, or impossible for a consumer app. Model the price at the scale you are hoping for, not the scale you start at.</p><h3>Where AI fits in mobile app development platforms</h3><p>The newest entrant changes the economics of question 7 and question 2 at the same time.</p><p>Low-code made building faster by replacing code with visual blocks, at the cost of ownership. AI app builders make building faster by writing the code for you, which means the speed and the ownership no longer have to trade off. The catch is that most AI app builders generate web apps. For mobile, the output has to be a real native project, or you are back to the web-view problem in question 1.</p><p>This is the gap we built <a href="https://www.rapidnative.com/?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=madp-guide&amp;utm_content=ai-section">RapidNative</a> to fill. You describe the app, or upload a sketch, a screenshot or a product requirements document, and it generates a full-stack React Native and Expo app: screens, navigation, a database, sign-in, file storage and realtime updates. You scan a QR code to run it on your own phone while you build, and change it by chatting or by clicking an element and describing the change. On paid plans you download the complete Expo project, written in TypeScript with Expo Router and NativeWind, with no proprietary runtime. When the app is ready, <a href="https://letsdeploy.it/?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=madp-guide&amp;utm_content=rapidnative-madp-guide">LetsDeploy.It</a>, our sister service, handles the App Store and Google Play submission.</p><p>Mapped against the six layers above, that covers front-end, backend, testing and publishing, with your own Supabase or APIs plugged in where you need them. Teams also use it alongside an existing stack: generate the first version in days, then move the exported project into their own repository and CI. Larger teams can bring their design system so generated screens match their brand.</p><p>It is not the right tool for everything. If you need to model hundreds of approval workflows across an ERP, an enterprise low-code platform with built-in process automation will serve you better. If you have a mature native team, a cross-platform framework and your existing CI may be all you need. AI-generated code is a starting point you own, not a replacement for engineering judgment on security and scale.</p><h3>Is MADP still the right term?</h3><p>Strictly, no. The analyst category moved on in 2019. But the need it named hasn’t changed: one supported, governed way to get good mobile apps built and shipped. If an RFP asks for a MADP, answer the underlying questions above rather than the acronym. For the full list of acronyms you will run into (MEAP, MBaaS, LCAP, MXDP, MDM, MAM and the rest), see our enterprise mobile glossary and our enterprise mobile app development guide on the RapidNative blog.</p><h3>The bottom line</h3><p>A mobile app development platform is less a product category than a promise: build once, ship to both stores, keep it running. Over fifteen years that promise has been sold as middleware, as cross-platform tooling, as low-code and now as AI. The best choice in 2026 is the one that gets a real native app on a real phone fastest, while leaving you holding source code you can take anywhere.</p><p>The quickest way to judge any platform is to build one real screen of your own app with it. If you want to try that with RapidNative, <a href="https://www.rapidnative.com/?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=madp-guide&amp;utm_content=free-plan">start from a prompt on the free plan</a>. Read the <a href="https://www.rapidnative.com/blogs/mobile-app-development-platform?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=madp-guide&amp;utm_content=end-cta">full guide on RapidNative</a> for links to everything mentioned here.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=896668b97483" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[You Have an App Idea. Here’s How to Turn It Into a Real App With AI.]]></title>
            <link>https://medium.com/@hugorus87/you-have-an-app-idea-heres-how-to-turn-it-into-a-real-app-with-ai-0fcc157d573e?source=rss-95a07a694494------2</link>
            <guid isPermaLink="false">https://medium.com/p/0fcc157d573e</guid>
            <category><![CDATA[app-idea]]></category>
            <category><![CDATA[android-app-development]]></category>
            <category><![CDATA[mobile-app-development]]></category>
            <dc:creator><![CDATA[Hugorus]]></dc:creator>
            <pubDate>Fri, 30 Oct 2026 13:03:43 GMT</pubDate>
            <atom:updated>2026-10-30T13:03:43.851Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A plain-English plan in four stages (plan, build, test, launch) for founders, small business owners and anyone who has never built an app before.</em></p><p>Almost everyone has an app idea sitting in their notes. A loyalty card for a favorite café. A way for a sports club to organize matches. A tracker for something nobody else seems to track.</p><p>For most people, that’s where the idea stays, because building an app used to mean hiring a developer or learning to code. That’s no longer true. Today, you can describe your app to an AI and hold a working version in your hand the same day.</p><p>What still decides whether your app succeeds isn’t the code. It’s the choices you make before and after the AI does its part. Here’s the plan we’d follow, in four stages.</p><p><em>Disclosure: we’re the team behind </em><a href="https://www.rapidnative.com/?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=idea-to-mobile-app-ai&amp;utm_content=intro"><em>RapidNative</em></a><em>, an AI app builder for mobile apps. The plan works whichever tool you use.</em></p><h3>Stage 1: Plan (30 minutes that save you days)</h3><h3>Decide what kind of app you’re making</h3><p>The first question isn’t about features. It’s this: <strong>do people need to download your app, or is a website enough?</strong></p><p>If your users will open it daily, want notifications or expect it on their home screen, you need a mobile app. If they’ll use it occasionally on a laptop, a web app may be fine.</p><p>This matters because many AI builders make web apps. They work nicely in a browser, but getting them into the App Store later usually means rebuilding them. If mobile is the goal, start with a builder that makes mobile apps from the beginning.</p><h3>Write your one-sentence app</h3><p>Take your idea and squeeze it into one sentence:</p><blockquote><strong><em>“My app helps [who] to [do one thing].”</em></strong></blockquote><p>Let’s use a running example: <em>“My app helps café regulars collect stamps for free coffee and see this week’s specials.”</em></p><p>If you can’t write that sentence, the AI can’t build a clear app either. And if your sentence has three “ands” in it, you’re describing three apps. Pick one.</p><h3>Name your screens</h3><p>Now list the screens someone needs to do that one thing. Usually it’s between three and six. For the café app:</p><ul><li><strong>Home:</strong> this week’s specials.</li><li><strong>Card:</strong> the stamp card, showing how many coffees until the free one.</li><li><strong>Scan:</strong> a QR code the barista scans to add a stamp.</li><li><strong>Profile:</strong> name, email and notification settings.</li></ul><p>That’s it. Everything else, like ordering ahead or a referral program, goes on a “later” list.</p><h3>Stage 2: Build (an afternoon)</h3><h3>Describe your app like you’d brief a designer</h3><p>AI app builders work best when you give them what a human designer would ask for: who it’s for, what it does, which screens it has and how it should feel. Here’s a prompt for the café app:</p><blockquote><em>Build a mobile app for a neighborhood café’s regular customers. Users collect a stamp for each coffee and get the 10th free. Screens: Home (this week’s specials with photos and prices), Card (a 10-stamp card showing progress), Scan (a large QR code for the barista to scan) and Profile (name, email, notification settings). Use a bottom tab bar. Style: warm and cozy, cream background, dark brown text, rounded cards and friendly illustrations.</em></blockquote><p>Paste it into <a href="https://www.rapidnative.com/?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=idea-to-mobile-app-ai&amp;utm_content=build">RapidNative</a>, and a few minutes later you’ll have a first version: real screens, navigation between them and the style you described, as a working React Native mobile app.</p><h3>Polish it with small requests</h3><p>Your first version is a draft. The fastest way to improve it is with one small, specific request at a time:</p><ul><li>Instead of <em>“make the card nicer”</em>, try <em>“make the stamps bigger and fill completed stamps with a coffee cup icon.”</em></li><li>Instead of <em>“fix the home screen”</em>, try <em>“show the price on the right of each special and make the photos square.”</em></li></ul><p>Work in this order: navigation first, then layout, then colors and details. Changing the structure late can undo your styling work.</p><h3>Stage 3: Test (a few days)</h3><h3>Hold it in your hand</h3><p>An app always feels different on a real phone. Buttons that looked big enough are suddenly hard to tap. Text that fit on your screen wraps on a smaller one. Use your app the way your customers would: standing at a counter, holding a coffee, in a hurry.</p><h3>Watch five real people use it</h3><p>Find five people who match your “who”. For the café app, that’s five regulars. Hand them the phone, ask them to do the main job, and say nothing. Every pause, every “where do I…?”, is something to fix.</p><p>This is the most valuable hour you’ll spend on your app, and it’s free.</p><h3>Connect real data</h3><p>Once the screens feel right, your app needs somewhere to keep real information, like user accounts and stamp counts. Most builders use a backend service such as Supabase or Firebase for that. With RapidNative you get standard React Native code, so you can connect the backend you prefer or export the code and continue with a developer.</p><h3>Stage 4: Launch (plan for two to three weeks)</h3><p>Launching takes longer than building, mostly because of store rules and review. Knowing them upfront prevents nasty surprises:</p><ul><li><strong>Developer accounts:</strong> Apple charges $99 a year, Google $25 once.</li><li><strong>Screenshots:</strong> your store listing sells your app before anyone downloads it. <a href="https://www.rapidnative.com/blogs/introducing-appshots-ai-app-store-screenshot-generator?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=idea-to-mobile-app-ai&amp;utm_content=appshots">AppShots</a> creates App Store screenshots with AI.</li><li><strong>A privacy policy</strong> that explains what data you collect.</li><li><strong>Account deletion</strong> inside the app, if people can sign up. Apple and Google both reject apps without it.</li><li><strong>A Google Play closed test:</strong> new personal developer accounts need 12 testers using the app for 14 days before it can go public.</li></ul><p>Then publish, tell your first users, and listen. The best version 2 is built from what real people ask for, not from your “later” list.</p><h3>What first-time builders ask us</h3><p><strong>Do I really not need to code?</strong> For designing and building your screens, no. For connecting a backend or publishing to the stores, a little technical help makes life easier, but many founders do it themselves with good guides.</p><p><strong>What if the AI gets it wrong?</strong> It will, sometimes. Treat it like a fast junior designer: give specific feedback, one change at a time, and it improves quickly.</p><p><strong>Can a developer take over later?</strong> Yes. RapidNative creates standard React Native and Expo code, which any React Native developer can work with.</p><p><strong>What should I build first?</strong> The smallest version that does your one main job well. If people love that, they’ll tell you what to add.</p><h3>The short version</h3><p>Write your one-sentence app. List the screens. Brief the AI like a designer. Polish with small requests. Test with five real people. Plan for the store rules before launch day.</p><p>Your idea has waited long enough. <a href="https://www.rapidnative.com/?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=idea-to-mobile-app-ai&amp;utm_content=conclusion">Start building it with RapidNative</a> and see it on your phone today.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0fcc157d573e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your App Has “Continue with Google.” Here’s Why Apple Rejected It.]]></title>
            <link>https://medium.com/@hugorus87/your-app-has-continue-with-google-heres-why-apple-rejected-it-e2fa8912f5a4?source=rss-95a07a694494------2</link>
            <guid isPermaLink="false">https://medium.com/p/e2fa8912f5a4</guid>
            <dc:creator><![CDATA[Hugorus]]></dc:creator>
            <pubDate>Thu, 29 Oct 2026 13:06:22 GMT</pubDate>
            <atom:updated>2026-10-29T13:06:22.414Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A plain-English guide to Guideline 4.8 for founders who built their app with Lovable, Bolt, Base44 or v0.</em></p><p>You built the app in a weekend. Login works, Google sign-in works, the web version has real users. You pay for an Apple Developer account, submit to the App Store, and two days later the reply arrives:\</p><blockquote><em>Guideline 4.8 — Design — Login Services</em></blockquote><p>No crash, no bug, no broken screen. Apple rejected your app because of the login screen your AI builder made for you.</p><p>Here’s what that rejection actually means, the three ways out, and which one to pick.</p><h3>Why did Apple reject my app for Google login?</h3><p>Because Google login shares information about your users with Google, and Apple wants iPhone users to have a more private choice on the same screen.</p><p>The rule is simple: if people can sign in with Google, Facebook, GitHub, X or any other outside account, your app must also offer a login option that:</p><ol><li>Only asks for a name and an email address.</li><li>Lets people hide their real email address, even from you.</li><li>Doesn’t track them for advertising without asking.</li></ol><p>Sign in with Apple does all three, which is why it’s the answer almost everyone ends up with.</p><p>AI app builders add Google login because it’s the easiest login to set up on the web. They don’t add Sign in with Apple, because a web app doesn’t need it. The rejection only shows up when the app goes to the App Store.</p><h3>You have three ways out</h3><h3>Option 1: Add Sign in with Apple (most apps)</h3><p>Keep Google login and add an Apple button above it. Your users keep the login they know, and iPhone users get a one-tap Face ID option.</p><p>Choose this if your users already sign in with Google, or if you want the fastest login on iPhone. It’s the option reviewers expect, so it’s the least likely to bounce back.</p><h3>Option 2: Remove social login (simplest apps)</h3><p>If your app only uses email and password, Guideline 4.8 doesn’t apply at all. Some founders take Google login out of the iOS app and keep email and password only.</p><p>Choose this if few people use Google login, or if you want the smallest possible change. The cost is a slower sign-up for new users.</p><h3>Option 3: Check whether you’re exempt (rare)</h3><p>The rule doesn’t apply to apps where people sign in with a school or work account, a government ID, or directly to the service the app is built for (like an app that manages your Google Calendar).</p><p>Choose this only if your app really is one of these. If you’re not sure, you’re probably not exempt.</p><h3>What does adding Sign in with Apple involve?</h3><p>For a typical AI-built app that stores users in Supabase, there are four pieces:</p><ul><li><strong>Apple Developer settings:</strong> switch on Sign in with Apple for your app.</li><li><strong>Supabase settings:</strong> turn on Apple as a login provider.</li><li><strong>The app itself:</strong> add Apple’s official button and connect it to Supabase. In a native React Native and Expo app, that’s a small package and about twenty lines of code.</li><li><strong>A test on a real iPhone:</strong> sign in with a fresh Apple ID and choose “Hide My Email”.</li></ul><p>If you’re doing it yourself, a prompt like this gets most builders started:</p><blockquote><em>“Add Sign in with Apple to the login screen, above Continue with Google, using Apple’s official button style. Connect it to Supabase Auth. Save the user’s name on first sign-in, because Apple only sends it once.”</em></blockquote><p>Then test it yourself. The builder will tell you it’s done whether or not it works on a real iPhone.</p><h3>Three surprises nobody mentions</h3><p><strong>Apple only tells you the person’s name once.</strong> The very first time someone signs in, Apple sends their name. After that, never again. If your app doesn’t save it right away, your profile screens say “Hello, “ forever.</p><p><strong>“Hide My Email” addresses can’t receive your emails by default.</strong> People who hide their email get an address like abc123@privaterelay.appleid.com. Your password reset and receipt emails won&#39;t reach them until you register your sending domain with Apple.</p><p><strong>Deleting an account now has one more step.</strong> Once you offer Sign in with Apple, Apple expects you to cancel the person’s Apple login when they delete their account. If you haven’t built account deletion yet, that’s its own rejection waiting to happen; <a href="https://letsdeploy.it/guides/app-rejected-account-deletion?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=guideline-4-8-sign-in-with-apple&amp;utm_content=account-deletion-guide">here’s how to fix it</a>.</p><h3>How to resubmit without another rejection</h3><p>Write a short note to the reviewer in App Store Connect. Reviewers test by hand, so tell them exactly where to look:</p><blockquote><em>“We added Sign in with Apple to the login screen, above Continue with Google. It only requests name and email and supports Hide My Email.”</em></blockquote><p>Make the Apple button at least as big and as high on the screen as the Google button, and use Apple’s standard button design. A tiny custom button at the bottom of the screen is a common reason for a second rejection.</p><p>While you’re in there, check the other rejections that come in pairs with this one. Apps rejected under 4.8 often hit <a href="https://letsdeploy.it/guides/lovable-app-rejected-guideline-4-2?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=guideline-4-8-sign-in-with-apple&amp;utm_content=guideline-4-2-guide">Guideline 4.2</a> in the same review, and our <a href="https://letsdeploy.it/guides/lovable-launch-checklist-and-costs?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=guideline-4-8-sign-in-with-apple&amp;utm_content=launch-checklist">launch checklist</a> covers everything else Apple and Google ask for.</p><h3>Quick answers</h3><p><strong>How long does it take to fix?</strong> For most AI-built apps, an afternoon of work, plus a new build and a new review. Most reviews finish within a day or two.</p><p><strong>Will this change my Android app?</strong> No. It’s an App Store rule. You can show the Apple button only on iPhone and leave your Android login as it is.</p><p><strong>Can I add the Apple button to my website too?</strong> Yes, and Supabase supports it there, but it’s optional. Apple only checks the iPhone app.</p><p><strong>Will people who signed up with Google lose their accounts?</strong> No. Adding a new login option doesn’t touch existing accounts. Google users keep signing in with Google.</p><p><strong>Do I need a paid Apple Developer account?</strong> Yes, but you already need one to publish on the App Store at all.</p><h3>The short version</h3><p>If your iPhone app has Google login, put a Sign in with Apple button above it, save the name on first sign-in, and tell the reviewer where it is. That’s usually all Guideline 4.8 needs.</p><p>If you’d rather not deal with the native build, Apple settings and resubmissions yourself, that’s the work we do at LetsDeployIt: we turn AI-built web apps (like ones from <a href="https://letsdeploy.it/lovable-to-app-store?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=guideline-4-8-sign-in-with-apple&amp;utm_content=lovable">Lovable</a>) into native apps and take them through App Store review. See <a href="https://letsdeploy.it/how-it-works?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=guideline-4-8-sign-in-with-apple&amp;utm_content=how-it-works">how it works</a> or <a href="https://letsdeploy.it/pricing?utm_source=medium&amp;utm_medium=referral&amp;utm_campaign=guideline-4-8-sign-in-with-apple&amp;utm_content=pricing">our plans</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e2fa8912f5a4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Every Coding Agent Needs Its Own Backend]]></title>
            <link>https://medium.com/@hugorus87/every-coding-agent-needs-its-own-backend-da3f218662a1?source=rss-95a07a694494------2</link>
            <guid isPermaLink="false">https://medium.com/p/da3f218662a1</guid>
            <category><![CDATA[lrs]]></category>
            <category><![CDATA[tinbase]]></category>
            <category><![CDATA[pglite]]></category>
            <category><![CDATA[supabase]]></category>
            <category><![CDATA[docker-alternative]]></category>
            <dc:creator><![CDATA[Hugorus]]></dc:creator>
            <pubDate>Mon, 12 Oct 2026 09:56:52 GMT</pubDate>
            <atom:updated>2026-10-12T09:56:52.839Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Parallel AI agents broke the “one shared dev database” habit. Here’s the setup that fixes it, without twelve containers per agent.</em></p><p>A year ago, a typical development setup had one developer, one branch at a time, and one local database. If you used Supabase, that database was supabase start: a stack of Docker containers that ran all day in the background.</p><p>That setup quietly assumed something that isn’t true any more: <strong>only one thing is changing the database at a time.</strong></p><p>Today it’s common to run several AI coding agents in parallel, each in its own git worktree, each working on a different feature. One is adding a favorites table. Another is rewriting the RLS policies on orders. A third is writing tests that sign up users and delete them afterward. And all three are pointed at the same local database.</p><p>You can probably guess how that goes.</p><h3>What goes wrong with a shared database</h3><p><strong>Migrations collide.</strong> Agent A’s migration adds a column. Agent B’s branch doesn’t know about it, resets the database to its own migration history, and Agent A’s tests start failing for reasons that have nothing to do with its code.</p><p><strong>Test data leaks between agents.</strong> One agent’s test signs up test@example.com. Another agent&#39;s test tries to sign up the same email and fails. Now an agent is &quot;fixing&quot; a bug that doesn&#39;t exist.</p><p><strong>Resets destroy other agents’ work.</strong> An agent that runs db reset to get a clean slate wipes the data every other agent was relying on.</p><p><strong>Nobody can tell whose change broke what.</strong> When something fails, the agent sees an error, not the history. It does what agents do: changes code until the error goes away, often in the wrong place.</p><p>None of these are agent bugs. They’re the same problems you’d get if three developers shared one local database. We just never had three developers on one laptop before.</p><h3>The fix: one backend per worktree</h3><p>The idea is simple. Every worktree gets its own database, with its own migrations, its own data, and its own port. Agents can reset, migrate, and seed freely, because nothing else is using it.</p><p>The problem has always been cost. The Docker-based Supabase stack runs about a dozen containers and uses roughly 1.4 GB of RAM before you’ve written a row. One of those is fine. Three, on a laptop that’s also running three agents, an editor, and a browser, isn’t.</p><p>That’s the gap single-process backends fill. I work on <a href="https://tinbase.dev/?utm_source=medium&amp;utm_medium=content&amp;utm_campaign=agent_backend_per_worktree&amp;utm_content=intro">tinbase</a>, an open-source Supabase-compatible backend that runs as one process with no Docker. On macOS and Linux it boots real Postgres 17 in about two seconds and uses around 59 MB of RAM. The official supabase-js works unchanged, and it reads your existing supabase/migrations folder.</p><p>At that size, one backend per worktree stops being a luxury.</p><h3>The setup</h3><h3>1. Give each worktree its own port</h3><pre>git worktree add ../app-favorites -b feature/favorites<br>cd ../app-favorites<br>npx tinbase start --port 54322</pre><p>Your main checkout keeps its backend on the default port, 54321. Each worktree gets the next one. tinbase keeps the database inside the worktree&#39;s folder, so each one has separate data automatically.</p><h3>2. Point the worktree’s app at its own backend</h3><p>In the worktree’s .env.local:</p><pre>EXPO_PUBLIC_SUPABASE_URL=http://127.0.0.1:54322<br>EXPO_PUBLIC_SUPABASE_ANON_KEY=&lt;same local key as before&gt;</pre><p>tinbase prints the same well-known local demo keys as supabase start, so only the port changes.</p><h3>3. Tell the agent which backend is its own</h3><p>Agents follow written instructions. Put this in the worktree’s agent instructions file (CLAUDE.md, AGENTS.md, or your tool&#39;s equivalent):</p><pre>## Local backend<br>- This worktree&#39;s backend runs at http://127.0.0.1:54322. Never use another port.<br>- To start fresh, run `npx tinbase db reset`. This only affects this worktree.<br>- New tables go in a new migration file in supabase/migrations/, with RLS enabled.<br>- Never use a hosted Supabase project or a service-role key for development.</pre><p>That last line matters more with agents than with people. An agent that can’t make a query work will try whatever makes the error go away, and a production key is the fastest way to make permission errors disappear.</p><h3>4. Let tests use a throwaway database</h3><p>For tests, you don’t need a server at all. Run them against an in-memory database that starts fresh every time:</p><pre>npx tinbase start --memory --port 54330</pre><p>Or use tinbase’s in-process API from your test runner, so each test file gets its own backend with nothing to clean up. Agents can run tests as often as they like without stepping on anything.</p><h3>5. Clean up with the worktree</h3><p>When the feature merges:</p><pre>git worktree remove ../app-favorites</pre><p>The backend’s data lived in the worktree folder, so it goes with it. No orphaned containers or volumes to hunt down later.</p><h3>Why this works better than the alternatives</h3><p>Approach Problem with parallel agents One shared local stack Migrations, resets, and test data collide One Docker stack per worktree Roughly 1.4 GB of RAM each; three agents can exhaust a laptop A hosted dev project Shared by everyone, network latency on every query, and real credentials in agent hands One cloud branch per worktree Works, but slow to create, needs a network, and costs money per branch One single-process backend per worktree About 59 MB each, starts in seconds, isolated by default</p><p>Cloud branching is a great fit for preview environments and for testing against production-shaped data. For the inner loop of agents writing and testing code all day, local and isolated is faster and safer.</p><h3>The limits, honestly</h3><p>tinbase is alpha, and it’s built for local development, CI, and prototypes, not production. A few things to know before you rely on it:</p><ul><li>It covers roughly 80% of the supabase-js surface. MFA, SSO, phone auth, and pgvector aren&#39;t supported yet.</li><li>It handles one writer at a time per instance. That’s fine for one agent per backend, which is the whole point of this setup.</li><li>Edge Functions with npm: or jsr: imports need bundling first.</li></ul><p>So keep a final check against the real Supabase CLI, or a hosted preview, before you deploy. The migration files are the same, so that check costs nothing extra to set up.</p><h3>The bigger shift</h3><p>For most of software history, “the development environment” meant one person’s machine, set up once and used all day. Agents are turning it into something closer to a disposable sandbox: created for a task, used hard for an hour, and thrown away.</p><p>Every piece of the stack has to fit that shape. Editors already do. Git worktrees do. The database has been the last heavy, shared, long-lived thing on the machine.</p><p>It doesn’t have to be. If your agents each get their own branch, they should each get their own backend too.</p><p><em>tinbase is open source (MIT). The source, the compatibility table, and the list of known gaps are on </em><a href="https://github.com/tinbase/tinbase?utm_source=medium&amp;utm_medium=content&amp;utm_campaign=agent_backend_per_worktree&amp;utm_content=cta"><em>GitHub</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=da3f218662a1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[You Built Your App in a Weekend with AI. Here’s What It Takes to Get It into the App Store.]]></title>
            <link>https://medium.com/@hugorus87/you-built-your-app-in-a-weekend-with-ai-heres-what-it-takes-to-get-it-into-the-app-store-aad9b7c40842?source=rss-95a07a694494------2</link>
            <guid isPermaLink="false">https://medium.com/p/aad9b7c40842</guid>
            <category><![CDATA[no-code]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[mobile-app-development]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[mobile-apps]]></category>
            <dc:creator><![CDATA[Hugorus]]></dc:creator>
            <pubDate>Mon, 12 Oct 2026 09:49:31 GMT</pubDate>
            <atom:updated>2026-10-12T09:49:31.849Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Why most AI-built apps are websites, what Apple and Google actually want, and the three honest paths from “it works in my browser” to “it’s on my phone”</em></p><p>It’s never been easier to build software. Describe an idea to Lovable, Bolt, Replit, or a dozen other AI tools, and by Sunday evening you have something that works: sign-up, a dashboard, data that saves. You send the link to friends. They love it.</p><p>Then someone asks the question that changes everything: <em>“Is it in the App Store?”</em></p><p>That’s where a lot of weekend projects stall. Not because the idea is wrong or the app is broken, but because of one fact most founders don’t learn until they try: <strong>most AI app builders make websites, not apps.</strong> A website that works beautifully on a phone is still not something Apple or Google will list as an app on its own.</p><p>This guide explains what that difference actually means, the three ways across it, and how to choose, without the jargon.</p><h3>Website or app? Here’s how to tell</h3><p>If you built with a tool that gives you a link you open in a browser, you have a <strong>web app</strong>. Lovable, for example, builds React web apps; its code is made to run in a browser, not as a native iPhone or Android app.</p><p>A <strong>native app</strong> is different. It’s installed from the App Store or Google Play, it runs directly on the phone, and it can use things like push notifications, the camera, Face ID, widgets, and offline storage the way people expect.</p><p>A quick test: if your app’s address starts with https:// and you open it in Safari or Chrome, it&#39;s a web app. That&#39;s not a problem. It&#39;s just a different starting point.</p><h3>Why Apple won’t just take your website</h3><p>Apple’s App Store guidelines include a section on <strong>minimum functionality</strong> (Guideline 4.2). In plain terms, Apple expects an app to offer something beyond a repackaged website: features, content, and an experience that feels like it belongs on the phone.</p><p>So the shortcut most people try first, putting their website inside a thin app “shell,” is exactly the kind of submission that guideline exists to stop. It doesn’t fail every time, but it’s the most common way a first submission comes back rejected.</p><p>Google Play is generally less strict about this, but it has its own hurdle for new developers: personal accounts created after November 2023 must run a closed test with at least 12 people for 14 days before the app can go public.</p><h3>Path 1: Make it installable from the browser (fastest, no store)</h3><p>Before thinking about stores at all, consider whether you need them yet.</p><p>A web app can be made <strong>installable</strong>: users tap “Add to Home Screen,” and it gets an icon and opens full screen, like an app. On iPhone, installed web apps can even receive push notifications.</p><p><strong>Good for:</strong> validating your idea with real users this week, with no review and no developer fees.</p><p><strong>Not good for:</strong> being found in the App Store, in-app purchases, or features that need deeper access to the phone. Many people also never discover the “Add to Home Screen” step.</p><h3>Path 2: Wrap your website in an app shell (quick, risky on iPhone)</h3><p>Tools like Capacitor take your existing web app and package it inside a native shell you can submit to the stores. You keep your code, and you can add some native features through plugins.</p><p><strong>Good for:</strong> Android, internal or business tools, and apps where you add genuinely native features (push, camera, offline mode) rather than just shipping the website as-is.</p><p><strong>The risk:</strong> on iPhone, a wrapped website with nothing extra is the classic Guideline 4.2 rejection. If you take this path, plan to add real native features before you submit, not after the first rejection.</p><h3>Path 3: Rebuild it as a real mobile app (most work, most reliable)</h3><p>The third path is to rebuild your app’s screens in a framework made for phones, most commonly <strong>React Native</strong> with <strong>Expo</strong>. Your backend (your database, sign-in, and business logic) usually stays the same; it’s the front end that gets rebuilt.</p><p>This used to mean starting over. With today’s AI tools, much less so. Your web app already defines every screen, flow, and piece of data, which is most of the hard thinking. Several AI builders now generate React Native apps directly, so a rebuild can take days rather than months.</p><p><strong>Good for:</strong> anything you want to grow: consumer apps, subscriptions, apps that rely on notifications or the camera, and anything that needs to feel at home on a phone.</p><p><strong>The trade-off:</strong> more upfront work, and a second codebase to maintain if you keep the web version too.</p><h3>How to choose</h3><p>If this sounds like you… Start with “I just want to know if people will use it.” Path 1: installable web app “It’s an internal tool or mostly for Android.” Path 2: wrap it, and add native features “I want a real App Store app people pay for.” Path 3: rebuild in React Native “I don’t know yet.” Path 1 now, Path 3 once you have users</p><p>The most expensive mistake isn’t picking the “wrong” path. It’s picking Path 2 for an iPhone consumer app, getting rejected, adding features under pressure, getting rejected again, and spending weeks in review that could have gone into a rebuild.</p><h3>Whichever path you take, the stores ask for the same things</h3><p>Getting the app built is half the job. Before either store will list it, you’ll also need:</p><ul><li><strong>Developer accounts:</strong> $99 per year for Apple, $25 once for Google. Identity checks can take days, so start early.</li><li><strong>A privacy policy</strong> at a public web address that matches what your app actually collects.</li><li><strong>Privacy details</strong> in App Store Connect and Google’s <strong>Data Safety form</strong>.</li><li><strong>Account deletion inside the app</strong>, if people can sign up.</li><li><strong>Screenshots</strong> that show the real app, with content in them, not empty screens.</li><li><strong>A demo account</strong> so reviewers can get past your sign-in screen.</li><li><strong>For Android:</strong> the 14-day closed test with 12 or more testers, if your Google account is new.</li></ul><p>None of it is hard. It’s just a lot of small steps, and AI app builders don’t do any of it for you.</p><h3>The honest timeline</h3><p>For a first app on brand-new accounts, a realistic plan looks like this:</p><ul><li><strong>Path 1:</strong> live this week.</li><li><strong>Path 2:</strong> two to four weeks, longer if Apple rejects the first submission.</li><li><strong>Path 3:</strong> one to two weeks to rebuild with modern tools, then two to four weeks for store setup and review, with Android’s 14-day test running in parallel.</li></ul><p>If someone promises your AI-built website will be “in the App Store by Friday,” ask which of these paths they mean.</p><h3>The bottom line</h3><p>AI made building the first version of an app dramatically faster. It didn’t change what Apple and Google expect from an app on their stores. Know which kind of app you have, choose the path that fits where you are, and start the store paperwork the same day you start building.</p><p>The app you built in a weekend is real progress. It’s just the beginning of the route to the store, not the end.</p><p><a href="https://letsdeploy.it/?utm_source=medium&amp;utm_medium=content&amp;utm_campaign=vibecoded_app_to_app_store"><em>LetsDeployIt</em></a><em> takes apps built with AI tools from “it works in my browser” to live in the App Store and Google Play, including converting web apps to React Native and handling the store submission end to end.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aad9b7c40842" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Docs Page That Lied]]></title>
            <link>https://medium.com/@hugorus87/the-docs-page-that-lied-f6217f6fa243?source=rss-95a07a694494------2</link>
            <guid isPermaLink="false">https://medium.com/p/f6217f6fa243</guid>
            <category><![CDATA[documentation]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[web-development]]></category>
            <dc:creator><![CDATA[Hugorus]]></dc:creator>
            <pubDate>Mon, 12 Oct 2026 09:32:06 GMT</pubDate>
            <atom:updated>2026-10-12T09:32:06.713Z</atom:updated>
            <content:encoded><![CDATA[<p><em>How one aspirational page made us audit our entire docs system, and the question we now ask about every dependency</em></p><p>The thing that made us rebuild our docs wasn’t a framework complaint. It was a page.</p><p>The page was called “AI Integration.” It had a nice sidebar entry. It described, in warm, confident prose, how AppLighter integrated with Claude Code and Cursor rules: generic AI tooling, generic developer workflows, the kind of docs page that sounds like every other starter kit’s docs page.</p><p>The problem: AppLighter isn’t a generic starter kit. It’s a catalog of production React Native templates. The AI Integration page described a feature we didn’t ship. It had been written aspirationally, in the hope that someday we would build it, and then published before the thing it described existed.</p><p>Nobody caught it for months. Nobody complained. The page just sat there, quietly teaching every reader something that wasn’t true.</p><p>When we found it, we had two choices: build the thing the page described, or delete the page. We deleted the page. And that one deletion turned into an audit of every file in the docs system, because if we’d been shipping docs that didn’t match the product, what else didn’t match?</p><h3>The audit was uncomfortable</h3><p>The honest answer, it turned out, was most of the scaffolding.</p><p>We had a right-hand table of contents on every page. We looked at the analytics. Nobody used it. We removed it.</p><p>We had rehype-slug and rehype-autolink-headings wired up because the table of contents needed heading anchors. Without the table of contents, no anchors. We removed them.</p><p>We had Contentlayer compiling our MDX into typed objects. Builds were 30 seconds longer than they needed to be. The code it replaced was 20 lines of gray-matter and fs.readFileSync. We removed Contentlayer.</p><p>We had considered adding Algolia search. Nobody had ever asked for it, and the docs were small enough to scan in one sidebar. We never added it.</p><p>By the end of the audit, the docs system had lost about 60% of its surface area and gained zero features anyone missed. The docs got faster to build, faster to load, faster to edit, and, most importantly, more honest.</p><h3>The useful heuristic</h3><p>Here’s the test I now apply to every piece of infrastructure in a small codebase:</p><p><strong>If I removed this today, who would notice tomorrow?</strong></p><p>If the answer is “nobody,” remove it.</p><p>If the answer is “me, because I’d have to debug the thing it was papering over,” ask whether the thing it was papering over is still a real problem.</p><p>If the answer is “a user, immediately and visibly,” keep it.</p><p>Most docs tooling fails that test. Most developers don’t realize it fails, because the tooling arrived as a dependency of something they actually wanted. You install a framework for one feature and inherit twenty.</p><h3>What stayed</h3><p><strong>MDX stayed.</strong> The one real benefit of MDX over plain Markdown, being able to drop a React component inline when you need a callout or a diff viewer, is worth the cost of the parser.</p><p><strong>Shiki stayed.</strong> Code blocks in the docs look exactly like code blocks in VS Code. That continuity matters, and it costs almost nothing.</p><p><strong>Static generation stayed.</strong> Every docs page is pre-rendered at build time. If a page is wrong in production, it’s because we merged it wrong, not because of a stale cache.</p><p><strong>The sidebar stayed, ordered by hand.</strong> Alphabetical sidebars work against readers. “Getting Started” needs to come before “Core Concepts,” even though G comes after C.</p><p>That’s four things. The whole docs system now runs on about 400 lines of code and a handful of dependencies. We could print the entire thing and read it in an afternoon. We did, in fact, print the entire thing and read it in an afternoon. That’s the point.</p><h3>What this isn’t</h3><p>This isn’t an argument against docs frameworks. If your docs are 500 pages, if you need translations, if you have non-developer writers, if you need versioning on day one, use the framework. Frameworks exist because those problems are real, and they solve them.</p><p>This is an argument against installing a framework when you have none of those problems. The hand-rolled version isn’t better in every direction. It’s smaller, which is better in exactly one direction: the direction of the next bug.</p><h3>The real lesson</h3><p>I don’t think the rebuild made the docs much better. The content is roughly the same. The styling is roughly the same. The reader’s experience is roughly the same.</p><p>What the rebuild changed was my relationship with the docs. I now know every file, every dependency, and every rendering path. When something breaks, I know where to look, because the surface area is small enough to hold in my head.</p><p>That isn’t a feature readers experience directly. But it’s the feature that decides whether docs stay honest over time, because honest docs need someone to notice when they stop being honest, and nobody notices anything in a codebase they don’t understand.</p><p>The AI Integration page sat there lying because the system around it was too complicated for anyone to audit. The rebuild wasn’t about frameworks. It was about making a system small enough that untrue pages couldn’t hide in it.</p><p><em>AppLighter sells production-ready Expo + Supabase app templates. </em><a href="https://www.applighter.com/apps?utm_source=medium&amp;utm_medium=content&amp;utm_campaign=docs_page_that_lied"><em>Browse the catalog</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f6217f6fa243" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Shipping an Expo App Without Fighting Your Native Modules]]></title>
            <link>https://medium.com/@hugorus87/shipping-an-expo-app-without-fighting-your-native-modules-2362c95fa9a3?source=rss-95a07a694494------2</link>
            <guid isPermaLink="false">https://medium.com/p/2362c95fa9a3</guid>
            <category><![CDATA[ios]]></category>
            <category><![CDATA[expo]]></category>
            <category><![CDATA[react-native]]></category>
            <category><![CDATA[mobile-app-development]]></category>
            <category><![CDATA[android]]></category>
            <dc:creator><![CDATA[Hugorus]]></dc:creator>
            <pubDate>Tue, 06 Oct 2026 12:08:10 GMT</pubDate>
            <atom:updated>2026-10-06T12:08:10.147Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A config-plugin discipline that makes every SDK upgrade boring, in the best way.</em></p><p>Most Expo projects I audit carry the same hidden tax: a few hand-edited files in ios/ and android/. A permission someone added in Xcode. A meta-data tag pasted into AndroidManifest.xml. A Gradle property from a Stack Overflow answer. Each one works today and breaks quietly on the next SDK upgrade.</p><p>The fix is a habit, not a rewrite: describe every native change in config, so it’s regenerated correctly every time. This is the discipline that separates Expo projects that upgrade in an afternoon from the ones stuck three SDKs behind.</p><h3>Why hand edits are a time bomb</h3><p>Modern Expo projects use <strong>Continuous Native Generation</strong>: app.json (or app.config.js) describes the app, and npx expo prebuild generates the ios/ and android/ folders from it. In many projects those folders aren&#39;t even committed; they&#39;re rebuilt on demand by EAS Build.</p><p>That only works if everything native lives in config. Edit Info.plist by hand and the next prebuild --clean throws your edit away. Sometimes you notice immediately. Sometimes you notice when the App Store rejects a build for a missing permission string, two weeks after the upgrade.</p><h3>Step zero: you might not need a plugin at all</h3><p>Before writing anything, check whether app.json already covers it. A lot of native configuration has first-class fields:</p><pre>{<br>  &quot;expo&quot;: {<br>    &quot;ios&quot;: {<br>      &quot;infoPlist&quot;: {<br>        &quot;NSMicrophoneUsageDescription&quot;: &quot;Your microphone is used to record voice notes.&quot;<br>      },<br>      &quot;entitlements&quot;: { &quot;aps-environment&quot;: &quot;production&quot; }<br>    },<br>    &quot;android&quot;: {<br>      &quot;permissions&quot;: [&quot;android.permission.RECORD_AUDIO&quot;]<br>    }<br>  }<br>}</pre><p>Many libraries also ship their own config plugin with options, so the setup is one line in plugins:</p><pre>{<br>  &quot;expo&quot;: {<br>    &quot;plugins&quot;: [<br>      [&quot;expo-camera&quot;, { &quot;cameraPermission&quot;: &quot;Your camera is used to scan receipts.&quot; }]<br>    ]<br>  }<br>}</pre><p>Check the library’s README for “config plugin” before writing your own. Autolinking also handles native dependencies (CocoaPods and Gradle) for most modules, so Podfile edits are rarely needed.</p><p><strong>Write a plugin when the change has no </strong><strong>app.json field</strong>: a meta-data tag inside &lt;application&gt;, a Gradle property, an AppDelegate tweak, or anything a library&#39;s docs tell you to paste into a native file.</p><h3>What a config plugin actually is</h3><p>A config plugin is a function that receives the Expo config and returns a modified one. It runs during prebuild. Expo provides &quot;mods&quot; (like withInfoPlist, withAndroidManifest, and withGradleProperties) that hand you the parsed native file to change, so you never manipulate XML or plist text by hand.</p><p>It sounds deeper than it is. Most useful plugins are 20 to 30 lines.</p><h3>Your first plugin: an SDK key in both native configs</h3><p>A common case: a third-party SDK needs its API key in Info.plist on iOS and as a &lt;meta-data&gt; tag inside &lt;application&gt; on Android, and it ships no config plugin. There&#39;s no app.json field for that meta-data tag, so this is plugin territory.</p><pre>// plugins/withAcmeSdk.js<br>const {<br>  withInfoPlist,<br>  withAndroidManifest,<br>  AndroidConfig,<br>} = require(&#39;expo/config-plugins&#39;);</pre><pre>const withAcmeSdk = (config, { apiKey } = {}) =&gt; {<br>  if (!apiKey) {<br>    throw new Error(<br>      &#39;withAcmeSdk: missing apiKey. Set ACME_API_KEY in your environment.&#39;<br>    );<br>  }</pre><pre>  // iOS: add the key to Info.plist<br>  config = withInfoPlist(config, (cfg) =&gt; {<br>    cfg.modResults.AcmeApiKey = apiKey;<br>    return cfg;<br>  });</pre><pre>  // Android: add &lt;meta-data&gt; inside &lt;application&gt;<br>  config = withAndroidManifest(config, (cfg) =&gt; {<br>    const app = AndroidConfig.Manifest.getMainApplicationOrThrow(cfg.modResults);<br>    AndroidConfig.Manifest.addMetaDataItemToMainApplication(<br>      app,<br>      &#39;com.acme.sdk.API_KEY&#39;,<br>      apiKey<br>    );<br>    return cfg;<br>  });</pre><pre>  return config;<br>};</pre><pre>module.exports = withAcmeSdk;</pre><p>Register it in app.config.js, passing the key from the environment:</p><pre>// app.config.js<br>export default ({ config }) =&gt; ({<br>  ...config,<br>  plugins: [<br>    ...(config.plugins ?? []),<br>    [&#39;./plugins/withAcmeSdk&#39;, { apiKey: process.env.ACME_API_KEY }],<br>  ],<br>});</pre><p>Run npx expo prebuild --clean and check the generated files. The key now survives every prebuild and every SDK upgrade, and a missing key fails the build instead of shipping a broken app.</p><p>Two details worth copying:</p><ul><li><strong>Plain JavaScript.</strong> Local plugins referenced from config are simplest as .js files. TypeScript plugins work too, but need a compile step or a TypeScript-aware loader.</li><li><strong>Use the helpers.</strong> addMetaDataItemToMainApplication updates an existing tag instead of adding duplicates, so running prebuild twice gives the same result.</li></ul><h3>Four patterns that pay off at scale</h3><p><strong>1. One plugin per feature, not per file.</strong> A withVoiceNotes plugin that sets the iOS microphone string, the Android permission, and the background-audio mode is easier to reason about than three plugins named after the files they touch. When you remove the feature, you remove one plugin.</p><p><strong>2. Keep environment logic in </strong><strong>app.config.js, and native changes in plugins.</strong> Different bundle identifiers, app names, or URL schemes per environment belong in app.config.js, driven by an environment variable like APP_VARIANT. Plugins take those values as options. That keeps plugins reusable and environments visible in one place.</p><p><strong>3. Fail loudly.</strong> If a plugin needs a key, a file, or a value, throw a clear error when it’s missing. A prebuild that silently produces a broken native project is the worst outcome, because you find out in review or production.</p><p><strong>4. Make plugins idempotent.</strong> Running prebuild twice must produce the same files. Check before adding, use Expo&#39;s helpers, and never blindly push into arrays.</p><h3>When a plugin is the wrong tool</h3><p>Config plugins modify configuration. They’re not meant for writing native features.</p><ul><li><strong>You need your own native code</strong> (a Swift or Kotlin module): create a local Expo module with npx create-expo-module --local. It lives in your repo, gets autolinked, and survives prebuild like any other dependency.</li><li><strong>Your plugin has grown past a couple of hundred lines</strong> and rewrites native source files with regex: you’re fighting the system. Consider a local module, or opting out of Continuous Native Generation and committing ios/ and android/ as source you maintain by hand.</li></ul><p>Opting out is a legitimate choice. It just means you own those folders, and every SDK upgrade becomes a manual merge.</p><h3>Find your hidden hand edits today</h3><p>Here’s a 15-minute audit for an existing project:</p><ol><li>Create a branch.</li><li>If ios/ and android/ are committed, run npx expo prebuild --clean.</li><li>Run git diff.</li></ol><p>Every difference is a native change that lives only in a hand-edited file. For each one, ask: is there an app.json field for this? A library plugin option? If not, write a plugin. Then commit, and stop tracking the generated folders if your workflow allows it.</p><p>Your next SDK upgrade will be the most boring one you’ve had.</p><h3>The template mindset</h3><p>Once you’ve written a few plugins, you’ll notice they’re reusable. The same voice-notes, maps, or analytics setup shows up in project after project. Teams that ship Expo apps regularly end up keeping a starter repo with these plugins already in place, so each new project begins with a clean, upgradeable baseline.</p><p>That’s the idea behind <a href="https://www.applighter.com/?utm_source=medium&amp;utm_medium=content&amp;utm_campaign=expo_config_plugins_2026">AppLighter</a>: production-ready Expo + Supabase templates where native configuration lives in config, so upgrades stay boring from the first commit.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2362c95fa9a3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Built a Fitness Wearable App in React Native. Here’s Where the Time Actually Went.]]></title>
            <link>https://medium.com/@hugorus87/i-built-a-fitness-wearable-app-in-react-native-heres-where-the-time-actually-went-2a6c0e6aa8d7?source=rss-95a07a694494------2</link>
            <guid isPermaLink="false">https://medium.com/p/2a6c0e6aa8d7</guid>
            <category><![CDATA[mobile-app-development]]></category>
            <category><![CDATA[react-native]]></category>
            <category><![CDATA[mobile-apps]]></category>
            <category><![CDATA[android-app-development]]></category>
            <dc:creator><![CDATA[Hugorus]]></dc:creator>
            <pubDate>Mon, 28 Sep 2026 13:21:41 GMT</pubDate>
            <atom:updated>2026-09-28T13:21:41.231Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Two months, three BLE bugs, and one lesson about not writing the boring 75%.</em></p><p>The first time I tried to build a fitness companion app for a heart-rate strap, I gave myself a weekend. Two months later, I shipped.</p><p>Not because the tools are missing. React Native has fine libraries for HealthKit, for Health Connect, for Bluetooth. It’s because a fitness wearable app has about thirty small integrations, and each of them has an edge case that will humiliate you in a one-star review.</p><p>Here’s where the time actually went, and what I’d do differently.</p><h3>The UI is not where the time goes</h3><p>I spent the first week polishing the workout list, the live session screen, the little SVG chart of my heart rate over time. Beautiful. Useless. I hadn’t started on the actual wearable yet.</p><p>If I were doing it again, I’d get the scaffold out of the way in a day and spend the eight weeks on the parts I actually can’t skip. That’s roughly what a colleague did: they generated the whole scaffold with <a href="https://www.rapidnative.com/?utm_source=medium&amp;utm_medium=content&amp;utm_campaign=how-to-build-fitness-wearable-companion-app-react-native">RapidNative</a>, got the six screens and a Supabase-backed data layer for free, and then spent their real time on the Bluetooth code.</p><blockquote><em>The scaffold is not what makes a fitness app hard. The wearable is.</em></blockquote><h3>Three Bluetooth bugs, in order</h3><p>Once I actually started on Bluetooth, three things bit me.</p><p><strong>Scanning.</strong> I ran a scan every time the app opened, because how else do you find the device. On iOS, this drained the phone battery from 100% to 40% in three hours. The fix is obvious in retrospect: scan once for pairing, remember the device ID, connect directly next time. Nothing in the react-native-ble-plx docs waves a flag about it.</p><p><strong>Reconnection.</strong> A heart-rate strap disconnects constantly. Sweat, distance, an unlucky wrist twist. My code treated disconnection as an error state and gave up. Real apps treat it as a state machine: disconnected → reconnecting → connected, with exponential backoff, that survives the app going to the background.</p><p><strong>Backgrounding.</strong> Android 14 introduced foreground service types, and health tracking is one of them. If your app keeps BLE alive with the screen off, you need the FOREGROUND_SERVICE_HEALTH permission in the manifest and the same service type declared in Play Console. Miss either and Google Play rejects the submission. I found this out at midnight the day before my launch.</p><h3>HealthKit and Health Connect are easier than you think</h3><p>I was scared of HealthKit. I shouldn’t have been. react-native-health and react-native-health-connect are twenty lines of setup and give you step count, workouts, heart rate, sleep. Both ship Expo config plugins now, so you don&#39;t touch Xcode.</p><p>The one gotcha: Google shut down the Google Fit developer APIs in 2025, and Health Connect is the replacement. If a tutorial tells you to install react-native-google-fit, close the tab.</p><h3>The best decision I made was not writing the scaffold</h3><p>About 75% of any fitness app is the same as every other fitness app. Tab bar. Onboarding. Permission flow. Session list. Chart card. Settings. Auth. A Postgres schema for sessions and samples. That’s not where the value is.</p><p>The value is the wearable code: the part that talks to your specific device, computes training load correctly, presents the data in a way a coach would recognize. That’s the other 25% of the codebase, and it deserves 100% of your attention.</p><p>Next time, I’ll generate the 75% and write the 25%.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2a6c0e6aa8d7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Read Every React Native Support Ticket We Got Last Year. Here’s What I Learned.]]></title>
            <link>https://medium.com/@hugorus87/i-read-every-react-native-support-ticket-we-got-last-year-heres-what-i-learned-8c636c97c595?source=rss-95a07a694494------2</link>
            <guid isPermaLink="false">https://medium.com/p/8c636c97c595</guid>
            <category><![CDATA[developer-tools]]></category>
            <category><![CDATA[react-native]]></category>
            <category><![CDATA[mobile-development]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Hugorus]]></dc:creator>
            <pubDate>Fri, 25 Sep 2026 09:07:50 GMT</pubDate>
            <atom:updated>2026-09-25T09:07:50.775Z</atom:updated>
            <content:encoded><![CDATA[<p>There’s a moment, if you sell developer tooling, when you realize your inbox is the most honest research instrument you own. Nobody files a support ticket to make you feel good. Nobody writes a five-star review at 3am. They file a ticket because their build has been red for six hours and their launch is Monday.</p><p>For twelve months I tagged every ticket that came through AppLighter. Not the refund requests. Not “does this template support SDK 57?” The real ones: a developer stuck at a specific line, in a specific file, at a specific hour of the night, unable to move.</p><p>By the end of the year I had a very clear picture of what React Native developers actually struggle with. It looks almost nothing like what the annual community surveys tell you.</p><p>Everyone loves to write about how React Native re-renders are hard, or how Redux is complicated, or how the New Architecture is a big deal. And they are, in the abstract. But nobody files a ticket about a re-render. They file a ticket about eas build --profile production dying deep inside Fastlane with the words &quot;unable to sign&quot; and no other context.</p><h3>The pattern I didn’t expect</h3><p>The most common problem, by an embarrassing margin, was that a developer’s first EAS Build failed. Not their tenth. Their first. Before they’d written any real code.</p><p>Almost always, one of three things:</p><ul><li>Their local eas-cli was a major version behind the SDK.</li><li>Their Apple developer credentials had never been added.</li><li>They picked the development profile but were building for a physical device, and the provisioning profile hadn&#39;t been generated yet.</li></ul><p>None of this is a React Native problem. All of it feels like one when you’re inside it.</p><p>The second most common was Supabase RLS policies silently returning empty arrays. The developer was signed in. They could see their user ID. They could see the row in the Supabase dashboard. But select * returned []. Every time.</p><p>The reason, once you know it, is boring. auth.uid() in the policy doesn&#39;t match the user_id column on the row. Or the client is sending the anon key instead of the user&#39;s JWT. Or the session expired. Every instance of this ticket had the same shape, &quot;I&#39;m signed in but I can&#39;t see my data,&quot; and the fix was always the same: your policy is working perfectly, it&#39;s just filtering out everything.</p><h3>The other eight</h3><p>The rest of the top ten, roughly in order: environment variables not making it into the production binary; Metro bundler cache poisoning; Info.plist rejections from App Store review; Android package name conflicts after renaming; push notifications silently dropped on Android; AI streaming responses stalling on device; Stripe webhooks not reaching Supabase; and voice transcription latency on mid-range Android.</p><p>Notice something? None of them are the framework. They’re the seam between the framework and the platform. The framework has stabilized. Everything around it is still wet cement.</p><h3>Why this matters if you sell templates</h3><p>If you’re a template vendor, this changes what you build. It’s tempting to spend your time on the beautiful UI. It’s boring to spend it making sure the FCM setup docs match the current Firebase console. But when a meaningful share of your customers are stuck on Android push, and the biggest share on their very first build, you know exactly where your labor should go.</p><p>We changed a lot in response. Preflight checks before the first build. A real troubleshooting doc per template. RLS policies baked into the initial migration so nobody has to invent them. A one-command Metro cache reset so people don’t have to google the incantation. Boring, boring, boring work, and each piece killed a category of tickets.</p><h3>What this means if you’re a React Native developer</h3><p>If you’re the developer, not the vendor, take this away: your instincts about what’s hard are wrong. What’s hard isn’t the JavaScript. What’s hard is the perimeter. Learn eas.json before you learn React.memo. Learn NSCameraUsageDescription before you learn Reanimated. Get the environment-variable propagation model right on day one.</p><p>And if you’re picking a template, pick one whose vendor reads these tickets and iterates on them. If a template’s last commit was six months ago, it’s not stable. It’s abandoned.</p><p>I don’t think this is a very glamorous conclusion. But it’s what the tickets say. And the tickets are the closest thing we have to the truth in a market full of screenshots.</p><p><em>The templates that came out of these tickets are at </em><a href="https://www.applighter.com/?utm_source=medium&amp;utm_medium=content&amp;utm_campaign=support_tickets_2026&amp;utm_content=footer"><em>AppLighter</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8c636c97c595" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>