<?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 @nkita on Medium]]></title>
        <description><![CDATA[Stories by @nkita on Medium]]></description>
        <link>https://medium.com/@ankitasaxena10214?source=rss-77fab7cd2c2c------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*vIWThCOoPq8PWmxePRy5fg.png</url>
            <title>Stories by @nkita on Medium</title>
            <link>https://medium.com/@ankitasaxena10214?source=rss-77fab7cd2c2c------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 06:33:55 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ankitasaxena10214/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[Your AI Agent Is Stuck in a Loop. Here Are 11 Ways to Fix It]]></title>
            <link>https://medium.com/codetodeploy/your-ai-agent-is-stuck-in-a-loop-here-are-11-ways-to-fix-it-e81fbf38e3cf?source=rss-77fab7cd2c2c------2</link>
            <guid isPermaLink="false">https://medium.com/p/e81fbf38e3cf</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[@nkita]]></dc:creator>
            <pubDate>Sat, 07 Nov 2026 17:13:48 GMT</pubDate>
            <atom:updated>2026-11-07T17:13:48.856Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lDx6-FAsI68rN8ziXocteg.png" /></figure><p><em>Simple fixes for one of the most frustrating bugs in agent development</em></p><p>You ask your coding agent to fix one failing test.</p><p>It runs the tests. Reads the file. Changes a line. Runs the tests again. Same error.</p><blockquote><em>💥 </em><strong><em>Master AI &amp; Tech Skills </em></strong><br> 🔥 Up to 5<strong>0% OFF</strong> — Limited-Time Offer<br><a href="https://trk.udemy.com/zz4NBO"><em>👉 </em><strong><em>Enroll Now &amp; Start Learning</em></strong></a></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ITogvtC6mF2BUF35Qsd9pg.png" /></figure><p>So it changes the line back. Runs the tests. Same error. Twenty minutes later, it’s still going, and your token bill keeps climbing.</p><p>Welcome to the agent loop. It’s one of the most common bugs in agent development. The good news? It’s usually easy to fix.</p><h3>Why agents get stuck</h3><p>Every agent runs the same cycle: think, act, observe, repeat. A loop happens when that cycle never reaches “done.”</p><p>There are usually three reasons. The agent doesn’t know it’s finished. It doesn’t understand what went wrong. Or it has forgotten what it already tried.</p><p>Every fix below targets one of these.</p><h3>1. Set a hard step limit</h3><p>This is your safety net. Cap the number of steps an agent can take, say 10 or 15. When it hits the cap, stop and return what you have.</p><p>Most agent frameworks have a setting for this. Use it. An agent with no limit is an open tab on your API bill.</p><h3>2. Define what “done” looks like</h3><p>“Fix the tests” is vague. Which tests? All of them? What if one was already broken? The agent has no idea when to stop.</p><p>“Make test_login pass without breaking any other test” is clear. Put finish criteria like this in your prompt.</p><h3>3. Keep tool results short and clear</h3><p>Sometimes the agent did find the answer. It just couldn’t see it.</p><p>The test runner returns 3,000 lines of logs. The real error is one line near the bottom, so the agent misses it and runs the tests again.</p><p>Trim tool results to what matters: the failing test’s name and its error message. A clean result is one the agent can act on.</p><h3>4. Write error messages the agent can use</h3><p>Many loops start with a vague error. A tool returns “Error,” and the agent retries the exact same call, hoping for a new result.</p><p>Make errors useful: “File not found: src/auth.py. Did you mean src/auth/login.py?” Now the agent knows exactly what to change.</p><h3>5. Give it a “finish” tool</h3><p>Sometimes the agent wants to stop but has no clear way to say so. Add an explicit tool like final_answer or task_complete. Calling it ends the run.</p><p>Stopping becomes an action, not a guess.</p><h3>6. Keep a log of past attempts</h3><p>In long runs, older steps get trimmed from context. The agent forgets it already tried a fix, so it tries the same fix again.</p><p>Keep a short running note of what was tried and what happened. Feed it back to the agent every turn.</p><h3>7. Use fewer, clearer tools</h3><p>Twenty tools with similar names confuse the model. It bounces between run_tests, execute_tests and test_runner, never sure which one is right.</p><p>Give the agent only the tools this task needs. Clear names, clear descriptions, no overlap.</p><h3>8. Plan first, then act</h3><p>Agents that jump straight into action tend to wander. Ask for a short plan first: read the error, find the cause, fix it, run the tests once.</p><p>A plan gives the agent a map. Without one, it just keeps walking.</p><h3>9. Add a progress check</h3><p>Every few steps, have the agent pause and ask itself one question: “Is the error any different from three steps ago?”</p><p>If not, it should change its approach or stop. This simple check catches a surprising number of loops.</p><h3>10. Set budgets for tokens and time</h3><p>Step limits alone aren’t enough. A single step can still burn thousands of tokens or hang on a slow build.</p><p>Add a token budget and a timeout. When either runs out, the agent wraps up and returns what it has.</p><h3>11. Let it ask for help</h3><p>Some tasks are truly stuck. Missing permissions, a broken dev environment, or a bug inside a library the agent can’t change.</p><p>Don’t let the agent spin. Give it a way to say “I need a human,” and pass along what it tried. One clear handoff beats a hundred useless retries.</p><h3>The bottom line</h3><p>Before you try any of these, turn on logging. Trace every thought, tool call and result. You can’t fix a loop you can’t see.</p><p>Then start with the cheap wins: a step limit, a clear definition of done, and better error messages. These three alone fix most loops.</p><p>Add the rest as your agent grows more complex.</p><p>Because a good agent isn’t one that never gets stuck. It’s one that knows when to stop.</p><p><em>What’s the strangest loop your agent has gotten stuck in? Share it in the comments.</em></p><h3>Thank you for being a part of the community</h3><p><em>Before you go:</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*d9QTaaaxboQP_gKSLedW_w.png" /></figure><p>👉 Be sure to <strong>clap</strong> and <strong>follow</strong> the writer ️👏<strong>️️</strong></p><p>👉 Follow us: <a href="https://medium.com/codetodeploy"><strong>Medium</strong></a></p><p><strong>Disclosure:</strong> This post includes affiliate and partnership links.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e81fbf38e3cf" width="1" height="1" alt=""><hr><p><a href="https://medium.com/codetodeploy/your-ai-agent-is-stuck-in-a-loop-here-are-11-ways-to-fix-it-e81fbf38e3cf">Your AI Agent Is Stuck in a Loop. Here Are 11 Ways to Fix It</a> was originally published in <a href="https://medium.com/codetodeploy">CodeToDeploy</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From “AI Inside Apps” to “AI That Connects Apps” ]]></title>
            <link>https://medium.com/codetodeploy/from-ai-inside-apps-to-ai-that-connects-apps-1a6e6b9ce4fc?source=rss-77fab7cd2c2c------2</link>
            <guid isPermaLink="false">https://medium.com/p/1a6e6b9ce4fc</guid>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[@nkita]]></dc:creator>
            <pubDate>Wed, 07 Oct 2026 14:54:32 GMT</pubDate>
            <atom:updated>2026-10-07T14:54:32.397Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ohqZTXmjuasMGs7Sh-SzHQ.png" /><figcaption>Apps are becoming tools for AI agents.</figcaption></figure><p>One of the most interesting shifts happening in AI right now isn’t just about better models.</p><p>It’s about AI getting the ability to work with our apps, tools, data, and workflows.</p><blockquote><em>💥 </em><strong><em>Master AI &amp; Tech Skills</em></strong><br> 🎓 Get Up to 50% OFF Premium Courses<br> ⏰ Limited-Time Offer<br><a href="https://trk.udemy.com/zz4NBO"><em>👉 </em><strong><em>Enroll Now &amp; Start Learning</em></strong></a></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ITogvtC6mF2BUF35Qsd9pg.png" /></figure><p>The recent direction from HP and Perplexity is a great example of this shift.</p><p>Instead of thinking of AI as a chatbot where we ask questions and receive answers, we’re moving toward AI agents that can understand an objective, use tools, interact with applications, and complete tasks.</p><p>And this has some big implications for <strong>Flutter developers</strong>.</p><h3>The old model</h3><p>For many apps, AI integration looks something like:</p><p><strong>User → Flutter App → LLM → Response</strong></p><p>For example:</p><blockquote><em>“Summarize this document.</em></blockquote><p>Useful, but relatively isolated.</p><h3>The emerging model</h3><p>The architecture is becoming more like:</p><p><strong>User → AI Agent → Tools / Apps / APIs → Actions → Result</strong></p><p>Now imagine saying:</p><blockquote><em>“Find my latest travel information, check my calendar, compare the options, and prepare everything for my trip.”</em></blockquote><p>The AI needs to interact with multiple systems.</p><p>That means AI needs access to <strong>capabilities, not just information.</strong></p><h3>Why this matters for Flutter developers</h3><p>This is where technologies and concepts such as:</p><ul><li>Flutter + Dart</li><li>LLM APIs</li><li>Tool / function calling</li><li>MCP (Model Context Protocol)</li><li>On-device AI</li><li>Multimodal AI</li><li>Agentic workflows</li><li>Platform integrations</li><li>Secure API architecture</li></ul><p>become increasingly important.</p><p>Flutter developers already think in terms of reusable components, services, APIs, and cross-platform abstractions.</p><p>The next step is exposing those capabilities in a way that AI agents can understand and use.</p><p>Imagine a banking app exposing:</p><p><strong>getTransactions()</strong></p><p>A travel app:</p><p><strong>searchFlights()</strong></p><p>A productivity app:</p><p><strong>createTask()</strong></p><p>An expense app:</p><p><strong>categorizeExpenses()</strong></p><p>The AI doesn’t need to understand the internal implementation of every application.</p><p>It needs to understand <strong>what capabilities are available and when to use them.</strong></p><h3>MCP makes the idea even more interesting</h3><p>MCP is emerging as a standardized way for AI systems to connect with external tools and data.</p><p>That means we can start thinking beyond:</p><p><strong>“How do I put an LLM into my Flutter app?”</strong></p><p>and instead ask:</p><p><strong>“How can my Flutter app become a capability that an AI agent can use?”</strong></p><p>That’s a fundamentally different mindset.</p><h3>The future Flutter developer?</h3><p>I think the skill set is gradually expanding from:</p><p><strong>Dart + Flutter + APIs + State Management + App Architecture</strong></p><p>toward:</p><p><strong>Dart + Flutter + AI Models + Tools + Agents + Context + Permissions + Evaluation</strong></p><p>And there’s an important architectural challenge here:</p><p><strong>What should happen on-device, and what should happen in the cloud?</strong></p><p>On-device AI can provide privacy, low latency and offline capabilities.</p><p>Cloud models can provide larger models and more complex reasoning.</p><p>The interesting applications will likely combine both.</p><h3>A project I’d build today</h3><p>Instead of another ChatGPT clone, I’d build:</p><p><strong>An AI-native Flutter application where an agent can understand a user’s request and interact with the app through tools, APIs and platform capabilities.</strong></p><p>For example:</p><blockquote><em>“Prepare my monthly expense report.”</em></blockquote><p>The agent could:</p><ol><li>Access transactions</li><li>Categorize expenses</li><li>Identify anomalies</li><li>Generate a summary</li><li>Create a report</li><li>Save/share the result</li></ol><p>That’s no longer simply <strong>AI inside an app.</strong></p><p>It’s an app designed to <strong>participate in an AI ecosystem.</strong></p><p>And I think that’s one of the most important changes Flutter developers should start preparing for.</p><p>The next generation of apps may not just have AI features.</p><p><strong>They may become tools that AI agents can use.</strong></p><h3>Thank you for being a part of the community</h3><p><em>Before you go:</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*d9QTaaaxboQP_gKSLedW_w.png" /></figure><p>👉 Be sure to <strong>clap</strong> and <strong>follow</strong> the writer ️👏<strong>️️</strong></p><p>👉 Follow us: <a href="https://medium.com/codetodeploy"><strong>Medium</strong></a></p><p>👉 CodeToDeploy Tech Community is live on Discord — <a href="https://discord.gg/ZpwhHq6D"><strong>Join now!</strong></a></p><p><strong>Disclosure:</strong> This post includes affiliate and partnership links.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1a6e6b9ce4fc" width="1" height="1" alt=""><hr><p><a href="https://medium.com/codetodeploy/from-ai-inside-apps-to-ai-that-connects-apps-1a6e6b9ce4fc">From “AI Inside Apps” to “AI That Connects Apps” 🚀</a> was originally published in <a href="https://medium.com/codetodeploy">CodeToDeploy</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Stopped Running My Flutter App Just to See One Widget — Here’s How]]></title>
            <link>https://medium.com/codetodeploy/i-stopped-running-my-flutter-app-just-to-see-one-widget-heres-how-1d647d4b31c9?source=rss-77fab7cd2c2c------2</link>
            <guid isPermaLink="false">https://medium.com/p/1d647d4b31c9</guid>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[flutter]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[productivity]]></category>
            <dc:creator><![CDATA[@nkita]]></dc:creator>
            <pubDate>Thu, 01 Oct 2026 20:09:36 GMT</pubDate>
            <atom:updated>2026-10-01T20:09:36.062Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9x3gw_6hA1sWLcvEBNW92A.png" /></figure><p>One of the things I like about Flutter is how quickly you can see your changes. Hot Reload makes UI development much faster, but there was still one thing that annoyed me.</p><blockquote><em>💥 </em><strong><em>Master AI &amp; Tech Skills</em></strong><br> 🎓 Get Up to 50% OFF Premium Courses<br> ⏰ Limited-Time Offer<br><a href="https://trk.udemy.com/zz4NBO"><em>👉 </em><strong><em>Enroll Now &amp; Start Learning</em></strong></a></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ITogvtC6mF2BUF35Qsd9pg.png" /></figure><p>I would make a small change to a widget, run the app, navigate to the screen where that widget was being used, and then check the result.</p><p>If the change wasn’t quite right, I’d go back, make another change, and do the same thing again.</p><p>It’s not a big deal when you do it once. But when you’re spending an hour working on a single component, it gets repetitive.</p><p>That’s where <strong>Widget Preview</strong> caught my attention.</p><h3>What if I only want to see one widget?</h3><p>Let’s say I’m building a profile card.</p><pre>class ProfileCard extends StatelessWidget {<br>  final String name;<br>  final String role;</pre><pre>  const ProfileCard({<br>    super.key,<br>    required this.name,<br>    required this.role,<br>  });</pre><pre>  @override<br>  Widget build(BuildContext context) {<br>    return Card(<br>      child: Padding(<br>        padding: const EdgeInsets.all(16),<br>        child: Column(<br>          crossAxisAlignment: CrossAxisAlignment.start,<br>          children: [<br>            Text(<br>              name,<br>              style: Theme.of(context).textTheme.titleLarge,<br>            ),<br>            const SizedBox(height: 4),<br>            Text(role),<br>          ],<br>        ),<br>      ),<br>    );<br>  }<br>}</pre><p>Normally, I would put this widget somewhere in my app and then run the application to see how it looks.</p><p>But what if I could just tell Flutter:</p><blockquote><em>“I want to see this widget.”</em></blockquote><p>That’s basically the idea behind Widget Previews.</p><p>You can create a small preview for the component:</p><pre>@Preview(name: &#39;Profile Card&#39;)<br>Widget profileCardPreview() {<br>  return const ProfileCard(<br>    name: &#39;Alex Johnson&#39;,<br>    role: &#39;Flutter Developer&#39;,<br>  );<br>}</pre><p>Now I have a dedicated place to look at my ProfileCard without having to build a whole screen around it.</p><p>And that’s the part I find useful.</p><h3>It’s especially useful when you’re working on UI</h3><p>Imagine you’re working on a button.</p><p>You change the padding.</p><p>Then the font.</p><p>Then the border radius.</p><p>Then you realize the text looks strange when it’s longer.</p><p>With the normal approach, you might keep going back and forth between your code and the application.</p><p>With a preview, you can create a few different examples and see them while working on the component.</p><p>For example, you could have a normal button and another one with a longer label.</p><pre>@Preview(name: &#39;Normal&#39;)<br>Widget normalButton() {<br>  return MyButton(<br>    text: &#39;Continue&#39;,<br>    onPressed: () {},<br>  );<br>}</pre><pre>@Preview(name: &#39;Long Text&#39;)<br>Widget longTextButton() {<br>  return MyButton(<br>    text: &#39;Continue to checkout&#39;,<br>    onPressed: () {},<br>  );<br>}</pre><p>This becomes pretty useful when your widget has several different states.</p><p>You can preview an empty state, a loading state, an error state, or just different types of content.</p><h3>This is not a replacement for Hot Reload</h3><p>I don’t think Widget Previews make Hot Reload less useful.</p><p>They solve different problems.</p><p>Hot Reload is great when you’re working on the application itself. You change something, reload, and immediately see the result in the context of your app.</p><p>Widget Preview is more focused.</p><p>You’re saying, “I’m working on this particular component right now, and I just want to see it.”</p><p>For reusable components, that’s a nice distinction.</p><h3>I can also see this being useful for larger projects</h3><p>Most Flutter applications eventually end up with their own collection of reusable widgets.</p><p>You might have custom buttons, cards, input fields, dialogs, loading indicators, empty states, and so on.</p><p>Usually, when someone wants to see one of those components, they have to find a screen where it’s being used.</p><p>With previews, the component itself can have examples of how it’s supposed to look.</p><p>That can make a project easier to understand too.</p><p>Someone opening ProfileCard doesn&#39;t have to search the entire project to find an example. The preview is right there.</p><h3>But don’t throw away your real app</h3><p>There’s one thing worth keeping in mind.</p><p>A widget looking good in a preview doesn’t mean it’s going to work perfectly in the real application.</p><p>Your actual app still has real data, different screen sizes, navigation, state management, lists, API responses, and all the other things that can affect a widget.</p><p>So I wouldn’t use Widget Previews instead of testing the actual app.</p><p>I’d use them before that.</p><p>Build the component.</p><p>Check its different states.</p><p>Fix the obvious UI problems.</p><p>Then put it into the real application and test it there.</p><h3>A small feature, but a useful workflow change</h3><p>That’s probably what I like most about Widget Previews.</p><p>It’s not some huge feature that completely changes Flutter development.</p><p>It’s just a small improvement to the way you work.</p><p>Instead of repeatedly starting your app just to check a single component, you can work on that component directly.</p><p>And when you’re building a UI-heavy Flutter application, saving a few seconds here and there can add up pretty quickly.</p><p>Sometimes the best developer tools aren’t the ones that completely change how you code.</p><p>They’re the ones that remove the little annoyances you didn’t realize were slowing you down.</p><h3>Thank you for being a part of the community</h3><p><em>Before you go:</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*d9QTaaaxboQP_gKSLedW_w.png" /></figure><p>👉 Be sure to <strong>clap</strong> and <strong>follow</strong> the writer ️👏<strong>️️</strong></p><p>👉 Follow us: <a href="https://medium.com/codetodeploy"><strong>Medium</strong></a></p><p>👉 CodeToDeploy Tech Community is live on Discord — <a href="https://discord.gg/ZpwhHq6D"><strong>Join now!</strong></a></p><p><strong>Disclosure:</strong> This post includes affiliate and partnership links.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1d647d4b31c9" width="1" height="1" alt=""><hr><p><a href="https://medium.com/codetodeploy/i-stopped-running-my-flutter-app-just-to-see-one-widget-heres-how-1d647d4b31c9">I Stopped Running My Flutter App Just to See One Widget — Here’s How</a> was originally published in <a href="https://medium.com/codetodeploy">CodeToDeploy</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[4 Years of Flutter Taught Me Something AI Can’t: Knowing What Not to Build]]></title>
            <link>https://medium.com/codetodeploy/4-years-of-flutter-taught-me-something-ai-cant-knowing-what-not-to-build-5cdec218ab9f?source=rss-77fab7cd2c2c------2</link>
            <guid isPermaLink="false">https://medium.com/p/5cdec218ab9f</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[@nkita]]></dc:creator>
            <pubDate>Thu, 24 Sep 2026 15:15:12 GMT</pubDate>
            <atom:updated>2026-09-24T15:15:12.728Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*37SpCSJk4H6IhfG0XAu6CA.png" /></figure><p><strong>After four years of building Flutter apps, I’ve learned that good development isn’t always about building more. Sometimes, it’s about knowing what to leave out.</strong></p><blockquote><em>💥 </em><strong><em>Master AI &amp; Tech Skills</em></strong><br> 🎓 Get Up to 50% OFF Premium Courses<br> ⏰ Limited-Time Offer<br><a href="https://trk.udemy.com/zz4NBO"><em>👉 </em><strong><em>Enroll Now &amp; Start Learning</em></strong></a></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ITogvtC6mF2BUF35Qsd9pg.png" /></figure><p>When I started with Flutter, I was excited about building almost everything. If someone suggested a new feature, my first thought was usually, “How can I build it?” I wanted to try new packages, create better animations, improve the UI, and experiment with different approaches.</p><p>At that stage, writing code felt like the main part of being a developer.</p><p>But working on real applications slowly changed that mindset.</p><p><strong>I started realizing that every feature we build becomes something we have to maintain.</strong></p><p>A feature doesn’t disappear after it is released. Someone has to test it, fix bugs, handle edge cases, support it on different devices, and update it when the application changes. Sometimes a feature that takes only a few hours to build can create problems for months or even years.</p><p>That’s when I started looking at development differently.</p><h3>Building Is Easy. Deciding What to Build Is Hard.</h3><p>Flutter already makes application development relatively fast. You can create a screen, connect an API, add state management, and have a working feature without spending weeks on the basics.</p><p>Now AI has made the process even faster.</p><p>I can ask AI to create a widget, generate Bloc code, write an API model, explain an error, or suggest different approaches to a problem. Things that once required hours of typing can sometimes be done in minutes.</p><p>But this creates an interesting problem.</p><p><strong>When it becomes easier to build something, it also becomes easier to build something unnecessary.</strong></p><p>Before AI, spending several hours implementing a feature naturally made you think twice about whether it was worth doing. Now the implementation itself can be very cheap. The harder question is no longer “Can I build this?”</p><p>The harder question is <strong>“Should I build this?”</strong></p><h3>Not Every User Problem Needs a New Feature</h3><p>One of the biggest lessons I’ve learned is that a product doesn’t become better simply because we keep adding features.</p><p>Imagine a user has a problem with an existing flow. Our first reaction might be to create another button, another screen, or another setting to solve it. But sometimes the real solution is to simplify the existing flow instead of adding something new.</p><p>Maybe the user doesn’t need another option.</p><p>Maybe the existing feature is just difficult to understand.</p><p>Maybe the information is already available somewhere else.</p><p>Or maybe the problem isn’t important enough to justify another piece of functionality.</p><p><strong>Sometimes the best feature is a simpler version of the feature you were originally planning to build.</strong></p><h3>Every Feature Comes With a Hidden Cost</h3><p>When we estimate a feature, we usually think about development time.</p><p>“Maybe it will take two days.”</p><p>But that’s only the beginning.</p><p>After those two days, there will be testing, bug fixes, UI adjustments, analytics, documentation, support, and future changes. If the feature depends on another package or service, that adds another layer of responsibility.</p><p>This is something you understand much better after working on applications for a few years.</p><p><strong>The cost of a feature isn’t just the time required to build it. It’s the responsibility you create by adding it.</strong></p><h3>AI Can Help Me Build Faster. It Can’t Decide What Matters.</h3><p>I use AI in development, and I think it is an incredibly useful tool.</p><p>It can save time on repetitive work and help me explore solutions much faster. If I know what I want to build, AI can often help me get there much quicker than I could alone.</p><p>But there is an important difference between <strong>building a solution</strong> and <strong>understanding whether the solution is necessary</strong>.</p><p>If I tell AI, “Build this feature,” it will focus on helping me build it. It doesn’t know everything about the users, the product history, the business priorities, or the problems the team has already faced.</p><p>That’s where human judgment still matters.</p><h3>Experience Changes the Questions You Ask</h3><p>When you’re new to development, your questions are mostly technical.</p><p>How do I create this screen? Which package should I use? Should I use Bloc? How should I structure this API? How can I fix this error?</p><p>After working on real projects, the questions start changing.</p><p>Why are we building this? Who actually needs it? What happens if we don’t build it? Is there a simpler solution? Will this make the application easier or harder to maintain?</p><p>That change in thinking is something I didn’t really understand when I started.</p><p><strong>Experience isn’t only about knowing more code. It’s about knowing when you don’t need more code.</strong></p><h3>Sometimes Saying “No” Is Good Engineering</h3><p>Developers don’t always get rewarded for removing things.</p><p>Adding a feature gives us something visible to show. Removing a feature can look like we didn’t do any work.</p><p>But sometimes saying no can save the team a lot of work.</p><p>If a feature has very little value but creates a lot of complexity, not building it can be the better engineering decision. If an existing solution already solves the problem, adding another solution may only confuse users and developers.</p><p>The difficult part is that these decisions aren’t always exciting.</p><p>There is no fancy UI to show.</p><p>There is no impressive code to review.</p><p>But the impact can be much bigger than another feature.</p><h3>AI Is Making This Skill More Important</h3><p>I think this becomes even more important as AI improves.</p><p>AI will continue making development faster. Soon, the difference between spending two days and two hours on a feature may become much smaller.</p><p>That means implementation speed will become less of a limitation.</p><p>Our ability to decide what deserves our time will become more important.</p><p><strong>When everyone can build faster, knowing what not to build becomes a real advantage.</strong></p><h3>What Four Years of Flutter Taught Me</h3><p>Four years of Flutter didn’t just teach me about widgets, state management, APIs, debugging, performance, or architecture.</p><p>It taught me to pause before writing code.</p><p>Today, before I start building something, I try to understand the problem first. I want to know who needs the feature, what problem it solves, whether something already solves it, and what maintaining it will look like later.</p><p>I don’t always get that decision right.</p><p>But I ask the question now.</p><p>And sometimes, after discussing the requirement, the answer is surprisingly simple:</p><p><strong>We don’t need to build this.</strong></p><p>That doesn’t mean we are doing less as developers.</p><p>It means we are spending our time more carefully.</p><p>AI can help us write code faster. It can help us debug faster. It can help us learn faster.</p><p>But deciding <strong>what deserves to exist</strong> is still a very human part of development.</p><p><strong>Maybe the next level of being a developer isn’t learning how to build everything.</strong></p><p><strong>Maybe it’s learning what isn’t worth building.</strong></p><h3>Thank you for being a part of the community</h3><p><em>Before you go:</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*d9QTaaaxboQP_gKSLedW_w.png" /></figure><p>👉 Be sure to <strong>clap</strong> and <strong>follow</strong> the writer ️👏<strong>️️</strong></p><p>👉 Follow us: <a href="https://medium.com/codetodeploy"><strong>Medium</strong></a></p><p>👉 CodeToDeploy Tech Community is live on Discord — <a href="https://discord.gg/ZpwhHq6D"><strong>Join now!</strong></a></p><p><strong>Disclosure:</strong> This post includes affiliate and partnership links.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5cdec218ab9f" width="1" height="1" alt=""><hr><p><a href="https://medium.com/codetodeploy/4-years-of-flutter-taught-me-something-ai-cant-knowing-what-not-to-build-5cdec218ab9f">4 Years of Flutter Taught Me Something AI Can’t: Knowing What Not to Build</a> was originally published in <a href="https://medium.com/codetodeploy">CodeToDeploy</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Chasing a Ghost Bug: How Duplicate API Calls Faked an Empty Screen]]></title>
            <link>https://blog.stackademic.com/chasing-a-ghost-bug-how-duplicate-api-calls-faked-an-empty-screen-c785556e1890?source=rss-77fab7cd2c2c------2</link>
            <guid isPermaLink="false">https://medium.com/p/c785556e1890</guid>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[flutter]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[@nkita]]></dc:creator>
            <pubDate>Thu, 24 Sep 2026 02:00:11 GMT</pubDate>
            <atom:updated>2026-09-24T02:00:11.069Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dS14PUTyBL3jRgKZ_gnS6A.png" /></figure><p>Every developer has one bug that haunts them. This one haunted us for days.</p><p>It started with a message from a user: “Your app says there are no race results. But I know there are — I can see them on the website.”</p><p>Simple complaint. Except it wasn’t simple at all.</p><h3>The Symptom That Made No Sense</h3><p>Our Flutter app shows race results in a long scrollable list. Users load more results as they scroll down — classic infinite scroll. One client kept reporting that partway through scrolling, the list would just stop. A “no results” message would appear, right in the middle of a list that clearly had more data.</p><p>Here’s the maddening part: we couldn’t reproduce it. Not once. Not on any of our test phones, not on any network we tried. It only happened on this one user’s device.</p><p>When a bug won’t show itself on your machine, you start doubting everything except your own code. That’s exactly where we went wrong.</p><h3>The Wrong Leads</h3><p>We chased three theories, in order:</p><p><strong>Theory 1: Backend bug.</strong> Maybe the server was sending bad or incomplete data. We looped in the backend team and asked them to check.</p><p><strong>Theory 2: Network issue.</strong> Maybe the user’s connection was dropping requests somewhere in transit.</p><p><strong>Theory 3: Bad data.</strong> Maybe there was some edge case in the race results themselves — a weird record that broke our parsing.</p><p>We spent real time on all three. And every time, the backend logs told us the same thing: the server had received the request and answered it correctly. There was no missing data. No error on their end. Nothing.</p><p>That’s the moment this stopped being a normal bug and started being a ghost.</p><h3>The Clue We Almost Missed</h3><p>Here’s the thing about “no results” screens — they’re supposed to mean exactly one thing: there’s no data. So naturally, we kept looking for reasons the data might be missing.</p><p>But an empty state and a failed request can look identical to a user. Nothing in our UI distinguished them. We were staring at the wrong end of the problem.</p><p>We finally decided to stop guessing and pull the actual error logs off the affected device — not the server logs, not our simulated tests, the real device logs from the real moment it broke.</p><p>And there it was.</p><p>Not a data problem. Not a network problem. A <strong>429 error</strong> — “too many requests.”</p><h3>What Was Actually Happening</h3><p>Our infinite-scroll logic had a subtle flaw. On slower networks, or when the scroll animation “bounced” at the edge of the list, our code could accidentally fire the <em>same</em> “load more” request multiple in a row, instead of once per scroll trigger.</p><p>Our server has a rate limiter designed to block exactly this kind of rapid-fire request spam from a single device. It was doing its job perfectly — catching the duplicate calls and rejecting them.</p><p>The problem was on our side. When the app received that rejection, it didn’t know how to handle it gracefully. Instead of showing a “something went wrong, try again” message, it fell back to the same generic empty state used when there’s genuinely no data left.</p><p>So a rate-limit error and an empty list looked <em>exactly</em> the same to the user. The app wasn’t broken. It was failing silently — and hiding the real problem behind a message that told a convincing lie.</p><h3>The Real Lesson</h3><p>This bug taught us something worth remembering: <strong>an empty screen doesn’t always mean empty data.</strong></p><p>Sometimes it means the app hit an error and quietly swallowed it, showing you the most harmless-looking message instead of the truth. From the outside, “no results” and “request failed” can be indistinguishable — unless you specifically design your app to tell them apart.</p><p>A few things we changed after this:</p><ul><li>Every failed request now shows its own distinct error state, never the generic empty message</li><li>We added debounce logic to stop duplicate scroll-triggered requests from firing</li><li>We now pull device-side logs much earlier when a bug won’t reproduce on our end</li></ul><p>The fix, once we found it, took an afternoon. Finding it took very much time of chasing shadows in the wrong direction — because we trusted the message on screen instead of questioning it.</p><p>Want this as a downloadable file (Markdown or Word doc) to publish or share elsewhere?</p><h3>Before you go</h3><ul><li>Please take a moment to like the post and follow the writer!</li><li>Did you know that over 400,000 developers share what they’re building, learning, and discovering across our platforms every month? Learn how you can contribute <a href="https://plainenglish.io/write-for-us"><strong>here</strong></a><strong>.</strong></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c785556e1890" width="1" height="1" alt=""><hr><p><a href="https://blog.stackademic.com/chasing-a-ghost-bug-how-duplicate-api-calls-faked-an-empty-screen-c785556e1890">Chasing a Ghost Bug: How Duplicate API Calls Faked an Empty Screen</a> was originally published in <a href="https://blog.stackademic.com/">Stackademic</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your Flutter App Doesn’t Need More Optimization — It Needs Less Work]]></title>
            <link>https://blog.stackademic.com/your-flutter-app-doesnt-need-more-optimization-it-needs-less-work-8a77da988fa7?source=rss-77fab7cd2c2c------2</link>
            <guid isPermaLink="false">https://medium.com/p/8a77da988fa7</guid>
            <category><![CDATA[flutter]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[coding]]></category>
            <dc:creator><![CDATA[@nkita]]></dc:creator>
            <pubDate>Wed, 16 Sep 2026 14:20:10 GMT</pubDate>
            <atom:updated>2026-09-16T14:20:10.214Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vKv1ATiux2Y00wTtodlypw.png" /></figure><p>Your app isn’t slow because Flutter is slow. It’s slow because it’s doing work nobody asked for.</p><p>I learned this the hard way. For years, every time an app felt laggy, I reached for the same toolbox: RepaintBoundary, const constructors, some clever caching. Those help. But they weren&#39;t fixing the real problem.</p><p><strong>The real problem was almost always memory — and images.</strong></p><p>Let me show you what I mean.</p><h3>The Silent Killer: Images</h3><p>Images are the #1 reason Flutter apps feel heavy. Not because Flutter is bad at rendering them. Because most apps load images the lazy way.</p><p>Here’s what usually happens:</p><ul><li>A list shows 50 profile pictures.</li><li>Each image is a 4000×3000 pixel photo straight from the camera.</li><li>Flutter decodes the <em>full</em> image into memory, even though it’s shown in a 60×60 pixel circle.</li></ul><p>That’s not a rendering problem. <strong>That’s 50 giant bitmaps sitting in RAM for no reason.</strong></p><h3>The Fix (and it’s simple)</h3><p>Don’t load the full image. Load only what you need.</p><ul><li>Use cacheWidth and cacheHeight on Image widgets to decode a smaller version.</li><li>Use the cached_network_image package — it caches images on disk <em>and</em> limits memory use.</li><li>Never fetch a 5MB image to show a thumbnail. Ask your backend for a resized version instead.</li></ul><p>One line of code — cacheWidth: 100 — can cut memory use for an image by 90%. That&#39;s not optimization. That&#39;s just not being wasteful.</p><h3>The Other Silent Killer: Memory Leaks</h3><p>Here’s a pattern I see in almost every Flutter codebase I review:</p><pre>class _MyWidgetState extends State&lt;MyWidget&gt; {<br>  final controller = TextEditingController();<br>  final scrollController = ScrollController();<br>  StreamSubscription? subscription;</pre><pre>  @override<br>  void initState() {<br>    subscription = someStream.listen((_) {});<br>  }<br>  // no dispose() method<br>}</pre><p>No dispose(). That means every time this widget is created and destroyed, its controllers, listeners, and subscriptions <strong>stay alive in memory</strong> — forever, until the app is killed.</p><p>Open that screen 20 times during a user’s session? You now have 20 ghost controllers doing nothing but eating RAM.</p><h3>The Fix</h3><p>Always pair creation with cleanup:</p><ul><li>Every AnimationController, TextEditingController, ScrollController → dispose it.</li><li>Every StreamSubscription → cancel it.</li><li>Every listener you add → make sure you remove it too.</li></ul><pre>@override<br>void dispose() {<br>  controller.dispose();<br>  scrollController.dispose();<br>  subscription?.cancel();<br>  super.dispose();<br>}</pre><p><strong>This one habit fixes more crashes and slowdowns than any performance library ever will.</strong></p><h3>Where Else Apps Do Unnecessary Work</h3><p>Once you start looking, you’ll find this pattern everywhere:</p><p><strong>1. Rebuilding the whole screen for one small change</strong> A single setState() at the top rebuilds 40 widgets to update one counter. Move state closer to what actually changes. Less gets rebuilt — not because you optimized it, but because you stopped asking for it.</p><p><strong>2. Holding onto things you don’t need anymore</strong> Static variables, global singletons, or long-lived services that hold a reference to a widget’s context — this quietly keeps that widget (and everything it touches) alive even after it&#39;s removed from the screen. Flutter&#39;s garbage collector can&#39;t clean up what your code is still holding onto.</p><p><strong>3. Doing heavy work on the main thread</strong> Parsing large JSON, resizing images, or sorting big lists directly in the UI thread blocks Flutter from drawing frames. Push it to an isolate — or better, ask if you even need to process that much data at once.</p><h3>A Real Example</h3><p>I once debugged an app that crashed after users browsed a product catalog for a few minutes. No obvious bug. No error logs pointing anywhere useful.</p><p>The cause: every product image was loaded at full resolution and never released. Scroll through 200 products, and you’ve loaded 200 full-size images into memory — with nothing evicting the old ones.</p><p><strong>The fix wasn’t a new caching library. It was three lines:</strong> resize the image before loading, and let the widget dispose naturally when it left the screen.</p><p>Crash gone. No optimization. Just less waste.</p><h3>The Real Mindset Shift</h3><p>Stop asking: <em>“How do I make this faster?”</em> Start asking: <strong>“What is my app holding onto that it doesn’t need anymore?”</strong></p><p>That question catches almost everything:</p><ul><li>Images loaded bigger than they’re shown.</li><li>Controllers that never got disposed.</li><li>Listeners nobody remembered to remove.</li><li>Data fetched and cached “just in case,” forever.</li></ul><p><strong>Every one of those is memory your app didn’t need to spend — and work it didn’t need to do.</strong></p><h3>Before You Reach for a Performance Fix</h3><p>Ask yourself:</p><ul><li>Is this image bigger than it needs to be?</li><li>Did I dispose everything I created in initState?</li><li>Is something still holding a reference to a screen that’s already gone?</li></ul><p>Most of the time, the fix isn’t clever. It’s just cleanup.</p><p><strong>A leaner app isn’t one that’s been optimized to death. It’s one that isn’t carrying weight it never needed in the first place.</strong></p><h3>Before you go</h3><ul><li>Please take a moment to like the post and follow the writer!</li><li>Did you know that over 400,000 developers share what they’re building, learning, and discovering across our platforms every month? Learn how you can contribute <a href="https://plainenglish.io/write-for-us">here</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8a77da988fa7" width="1" height="1" alt=""><hr><p><a href="https://blog.stackademic.com/your-flutter-app-doesnt-need-more-optimization-it-needs-less-work-8a77da988fa7">Your Flutter App Doesn’t Need More Optimization — It Needs Less Work</a> was originally published in <a href="https://blog.stackademic.com/">Stackademic</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Writes My Code Now. Here’s What I Actually Do With My Time.]]></title>
            <link>https://blog.stackademic.com/ai-writes-my-code-now-heres-what-i-actually-do-with-my-time-a7387fd82e75?source=rss-77fab7cd2c2c------2</link>
            <guid isPermaLink="false">https://medium.com/p/a7387fd82e75</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[coding]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[@nkita]]></dc:creator>
            <pubDate>Sun, 13 Sep 2026 11:42:37 GMT</pubDate>
            <atom:updated>2026-09-13T11:42:37.059Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ut6vDlsX9OiEqI71xj7puA.png" /></figure><p>Last month, I watched AI write an entire feature in under two minutes — a feature that would’ve taken me a full day. I sat there for a second, staring at the screen, wondering: what am I even for?</p><p>If you’re a developer in 2026, you’ve probably felt this too. The code writes itself now. So what’s left for us?</p><p>I spent weeks figuring this out. Here’s what I found.</p><h3>Speed Was Never the Real Skill</h3><p>I used to think my value was in how fast I could type code. Turns out, that was never it.</p><p>AI can write a function in seconds. But it can’t tell you if that function is solving the right problem. That’s where I come in.</p><h3>What I Actually Do Now</h3><p><strong>1. I ask better questions.</strong></p><p>Before AI writes a single line, I decide what needs to be built — and why. “Add a login screen” gets you generic code. “Add a login screen that remembers guest users for 30 days and syncs with our existing auth service” gets you something usable.</p><p>The prompt is the new spec. Writing a good spec is a skill on its own.</p><p><strong>2. I read code like a reviewer, not a writer.</strong></p><p>AI-generated code looks confident. That’s the trap — it can be clean, well-formatted, and still wrong.</p><p>Now I read every line like I’m reviewing a junior dev’s PR: checking edge cases, checking assumptions, checking for the one bug hiding in plain sight.</p><p><strong>3. I catch what AI can’t see.</strong></p><p>AI doesn’t know your users. It doesn’t know your app crashes on old Android phones in low-network areas. It doesn’t know your client hates loading spinners.</p><p>I do. That messy, real-world context still lives in my head, not in the model.</p><p><strong>4. I own the architecture.</strong></p><p>AI is great at writing one function. It’s not great at deciding how forty functions should talk to each other across a growing app.</p><p>That big-picture thinking is still mine. I decide the structure. AI fills it in.</p><p><strong>5. I test for the unexpected.</strong></p><p>AI tests for what it expects to happen. I test for what happens when the user loses internet mid-upload, rotates their phone at the wrong second, or taps a button twice by accident.</p><p>Real users are chaotic. AI hasn’t met them yet.</p><h3>The Part Nobody Talks About</h3><p>Here’s what surprised me most: I’m not coding less because I’m lazy. I’m coding less because I’m finally doing the parts that actually need a human.</p><p>For years, I confused “writing code” with “being a developer.” They’re not the same thing. Writing code was just the visible part — the thinking behind it was always the real job.</p><h3>So What Should You Do While AI Writes the Code?</h3><p>Here’s my honest list:</p><ul><li>Get better at describing problems, not just solving them</li><li>Read AI’s code the way you’d read a stranger’s — carefully</li><li>Keep the messy human context AI doesn’t have</li><li>Own the big picture, even when AI owns the small pieces</li><li>Test like a paranoid user, not a confident developer</li></ul><p>None of this shows up in a tutorial. But it’s the actual job now.</p><h3>The Shift Nobody Warned Us About</h3><p>I didn’t become a worse developer when AI started writing my code. I became a different one.</p><p>Less typing. More thinking. Less “how do I write this” and more “should this even exist.”</p><p>And honestly? That’s a better use of my time. AI took the part of the job that felt like work. What’s left is the part that actually needed a developer all along.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a7387fd82e75" width="1" height="1" alt=""><hr><p><a href="https://blog.stackademic.com/ai-writes-my-code-now-heres-what-i-actually-do-with-my-time-a7387fd82e75">AI Writes My Code Now. Here’s What I Actually Do With My Time.</a> was originally published in <a href="https://blog.stackademic.com/">Stackademic</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your Flutter App Isn’t Slow Because Flutter Is Slow — You’re Probably Doing This]]></title>
            <link>https://blog.stackademic.com/your-flutter-app-isnt-slow-because-flutter-is-slow-you-re-probably-doing-this-184aa6838102?source=rss-77fab7cd2c2c------2</link>
            <guid isPermaLink="false">https://medium.com/p/184aa6838102</guid>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[flutter]]></category>
            <category><![CDATA[mobile-app-development]]></category>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[@nkita]]></dc:creator>
            <pubDate>Tue, 08 Sep 2026 10:13:56 GMT</pubDate>
            <atom:updated>2026-09-08T10:13:56.103Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8tdI8ukACJesRFlvPz1sUw.png" /></figure><p>Flutter gets blamed a lot.</p><p>“It’s laggy.” “It’s not native enough.” “Maybe we should’ve gone with Kotlin.”</p><p>I’ve heard all of it. I’ve said some of it myself, back when I didn’t know better.</p><p>Then I actually profiled the app.</p><p>Turns out Flutter wasn’t the problem. <strong>I was.</strong></p><p>Here’s what’s actually slowing down most Flutter apps — and none of it is Flutter’s fault.</p><h3>1. You’re calling setState() like it’s free</h3><p>It’s not.</p><p>Every setState() rebuilds the entire widget subtree under it. Not just the button you tapped. Not just the text that changed.</p><p>Everything.</p><p>If your setState() sits at the top of a 40-widget screen, you just rebuilt 40 widgets to change one number.</p><p><strong>The fix:</strong> push state down. Wrap only the part that actually changes in its own widget, and call setState() there — not three levels up.</p><p>Small state, small rebuilds. That’s the whole trick.</p><h3>2. Your ListView is building things nobody’s looking at</h3><p>ListView(children: [...]) builds every single item immediately. All of them. Even the 400 that are off-screen.</p><p>Your users are looking at 6 items and your app is rendering 500.</p><p><strong>The fix:</strong> use ListView.builder(). It builds items lazily, only as they scroll into view.</p><p>This one change alone has fixed more “janky list” complaints than anything else I’ve tried.</p><h3>3. You’re judging performance in debug mode</h3><p>This one’s sneaky.</p><p>Debug mode ships with extra instrumentation, assertions, and a JIT compiler that’s built for hot reload — not speed. It is <em>supposed</em> to feel a little slower.</p><p>I’ve seen people rip out entire widgets because they felt “sluggish” in debug mode — when the actual release build ran perfectly fine.</p><p><strong>The fix:</strong> always benchmark in profile or release mode.</p><pre>flutter run --profile</pre><p>Judging Flutter’s speed from debug mode is like judging a car’s top speed with the handbrake on.</p><h3>4. Your images are doing way more work than they need to</h3><p>A 4000×3000 pixel photo, displayed in a 100×100 avatar, is still a 4000×3000 pixel image being decoded in memory.</p><p>Flutter doesn’t magically shrink it for you. It decodes the whole thing, every time, unless you tell it not to.</p><p><strong>The fix:</strong></p><ul><li>Use cacheWidth / cacheHeight to decode at the actual display size</li><li>Use cached_network_image so you&#39;re not re-downloading and re-decoding on every rebuild</li><li>Compress images before they ever reach the app</li></ul><p>I’ve seen scroll jank disappear entirely just from fixing image sizing. No fancy optimization — just decoding less.</p><h3>5. Your state management is rebuilding the whole tree</h3><p>Provider, Riverpod, Bloc — doesn’t matter which one. If you’re not scoping your listeners properly, you’re rebuilding way more than you think.</p><p>Wrapping your entire screen in one big Consumer means one small state change repaints the <em>entire</em> screen.</p><p><strong>The fix:</strong> scope your Consumer / Selector / BlocBuilder tightly, around just the widget that needs that piece of state. Not the screen. The widget.</p><p>Precision beats convenience here, every time.</p><h3>6. You’re parsing JSON on the main thread</h3><p>Small JSON payload? Fine, no one will notice.</p><p>Large payload — a big list, a nested API response — parsed with jsonDecode() directly on the main thread? That&#39;s a frozen frame waiting to happen.</p><p>The UI thread is busy decoding your JSON instead of drawing your next frame.</p><p><strong>The fix:</strong> move heavy parsing to an isolate with compute().</p><p>Your UI thread’s only job is drawing frames smoothly. Don’t make it do data work too.</p><h3>7. You forgot const</h3><p>This one feels too small to matter. It matters.</p><p>A const widget tells Flutter: &quot;this will never change, don&#39;t even bother rebuilding it.&quot; Without const, Flutter rebuilds it every single time its parent rebuilds — even if nothing about it actually changed.</p><p><strong>The fix:</strong> add const everywhere the widget is genuinely static. Icons, static text, spacing widgets — all of it.</p><p>It’s a one-word fix that costs you nothing and saves Flutter from doing pointless work, frame after frame.</p><h3>So… is Flutter slow?</h3><p>No.</p><p>Flutter can hit a rock-solid 60 or 120fps. Companies like Google Pay, Alibaba, and BMW ship performance-critical apps on it every day.</p><p>What’s usually slow is:</p><ul><li>Unscoped state updates</li><li>Eager list building</li><li>Oversized, uncompressed images</li><li>Main-thread work that should’ve been on an isolate</li><li>Missing const</li><li>Judging speed in the wrong build mode</li></ul><p>None of that is Flutter’s fault. All of it is fixable in an afternoon.</p><p><strong>If your app feels slow, don’t blame the framework — profile it.</strong> Nine times out of ten, the fix is smaller than you think, and it’s already on this list.</p><p>If you’ve hit a Flutter performance issue that <em>isn’t</em> on here, I’d love to hear about it — drop it in the comments.</p><h3>Before you go</h3><ul><li>Please take a moment to like the post and follow the writer!</li><li>Did you know that over 400,000 developers share what they’re building, learning, and discovering across our platforms every month? Learn how you can contribute <a href="https://plainenglish.io/write-for-us">here</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=184aa6838102" width="1" height="1" alt=""><hr><p><a href="https://blog.stackademic.com/your-flutter-app-isnt-slow-because-flutter-is-slow-you-re-probably-doing-this-184aa6838102">Your Flutter App Isn’t Slow Because Flutter Is Slow — You’re Probably Doing This</a> was originally published in <a href="https://blog.stackademic.com/">Stackademic</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Cut My Dev Time in Half — Not with AI, with This One Habit]]></title>
            <link>https://blog.stackademic.com/i-cut-my-dev-time-in-half-not-with-ai-with-this-one-habit-ab3d44fdc607?source=rss-77fab7cd2c2c------2</link>
            <guid isPermaLink="false">https://medium.com/p/ab3d44fdc607</guid>
            <category><![CDATA[mobile-app-development]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[@nkita]]></dc:creator>
            <pubDate>Sat, 05 Sep 2026 13:26:36 GMT</pubDate>
            <atom:updated>2026-09-05T13:26:36.550Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tQOdnsvMTNs3b6-s1kiLyw.png" /></figure><p>I Cut My Dev Time in Half — Not with AI, with This One Habit</p><p>I used to build first, think later. Open the IDE, start prompting, let the suggestions roll in, figure it out as I go.</p><p>For some days, that was my process. It felt productive — code appearing on screen, fast, like magic.</p><p>It wasn’t.</p><p>Three months ago, a project nearly broke me. A Flutter app, simple scope on paper. Three weeks turned into six.</p><p>Every feature I “finished” came back broken somewhere else. I’d fix one screen. Break another.</p><p>Sound familiar?</p><p>I was debugging more than I was building. Late nights, empty coffee cups, that specific exhaustion of chasing a bug you created yourself.</p><p>One night it got almost funny. I fixed a state management bug in the checkout screen. Two hours later, the profile screen broke — because of the exact same fix.</p><p>I stared at my screen and thought: I don’t even know why this app works when it works.</p><p>I didn’t need more hours. I needed a different habit.</p><p>The Habit</p><p>Before I touch the code now, I write the plan. Not documentation, not a spec doc nobody reads. Just five minutes, notes app, pen, whatever’s closest.</p><p>What am I actually building? What’s the simplest version that works? What could break?</p><p>That’s it. Five minutes.</p><p>The first time I tried it, I almost skipped it. Felt like a waste of time — I wanted to code, not write. I did it anyway, out of desperation more than discipline.</p><p>That feature took 40 minutes. The version before it, same complexity, had taken three hours.</p><p>I thought it was luck.</p><p>It wasn’t.</p><p>I tested it on the next ten features. Every single one, faster. Fewer bugs. Less backtracking.</p><p>My dev time didn’t drop by 10%. It dropped by half.</p><p>Why It Actually Works</p><p>Here’s what I didn’t understand before: coding isn’t the hard part. Thinking is the hard part.</p><p>When you code without a plan, you’re doing both at once — figuring out the logic and writing the implementation, at the same time. Your brain context-switches every few seconds.</p><p>That’s where the hours disappear.</p><p>Separate the two. Think first, fully, on paper. Then code — just translate, don’t invent.</p><p>It’s the difference between building a house with a blueprint, and building one brick by brick, hoping it works out.</p><p>Take that checkout bug. If I’d spent five minutes mapping how the state flowed across screens, I’d have seen the shared dependency instantly.</p><p>Instead, I found it the expensive way. Twice.</p><p>Before and After</p><p>Before: open IDE, start prompting, get stuck, backtrack, repeat. After: five minutes of planning, then a straight line to done.</p><p>Before: fixing bugs I didn’t remember creating. After: catching them on paper, before they became code.</p><p>Same skills. Same person. Different order of operations.</p><p>It’s Not Just a Flutter Thing</p><p>This isn’t a coding trick — it just happened to save my coding life. A friend uses it for writing emails now. My sister uses it for cooking, plans the whole meal before touching a pan.</p><p>The habit is the same everywhere: pause, plan, then execute.</p><p>We skip this because planning feels slower. It’s not — it just feels that way because you can see yourself not-coding.</p><p>You can’t see the three hours you’re about to save.</p><p>What Actually Changed</p><p>I ship features faster now. But that’s not the real win.</p><p>The real win: I stopped dreading my own code. I don’t open old files scared of what I’ll find. I don’t lie awake wondering what broke.</p><p>Everyone’s chasing tools to code faster — AI autocomplete, smarter IDEs, better frameworks. I use those too, and they help.</p><p>But none of them fixed the actual problem. The problem was never how fast I could type, or prompt. It was how much I was thinking while building.</p><p>The biggest unlock wasn’t a tool. It was five minutes of silence before I opened the IDE at all.</p><p>Try it on your next task.</p><p>Before you open your editor — stop. Write three lines: what, how, what could break. Then build.</p><p>You’ll code slower for five minutes. And ship twice as fast for the rest of the day.</p><p>No new tool required. Just a pause you were always allowed to take.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ab3d44fdc607" width="1" height="1" alt=""><hr><p><a href="https://blog.stackademic.com/i-cut-my-dev-time-in-half-not-with-ai-with-this-one-habit-ab3d44fdc607">I Cut My Dev Time in Half — Not with AI, with This One Habit</a> was originally published in <a href="https://blog.stackademic.com/">Stackademic</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Day I Realized I Was Coding With AI, Not Just Using It]]></title>
            <link>https://blog.stackademic.com/the-day-i-realized-i-was-coding-with-ai-not-just-using-it-9ff254979161?source=rss-77fab7cd2c2c------2</link>
            <guid isPermaLink="false">https://medium.com/p/9ff254979161</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[innovation]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[@nkita]]></dc:creator>
            <pubDate>Thu, 03 Sep 2026 11:51:28 GMT</pubDate>
            <atom:updated>2026-09-03T11:51:28.917Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dw8grcL0bdw0ddPMHZ6jnw.png" /></figure><h3>I thought I was using a tool. Turns out, I’d picked up a collaborator.</h3><p>It happened on a Tuesday. Nothing special about the day — just another sprint, another feature, another cup of chai going cold next to my laptop.</p><p>I was building a small AI-powered chat screen inside a Flutter app. Nothing fancy. I opened my editor, typed half a sentence describing what I wanted, and before I could finish my thought — the suggestion was already there.</p><p>Not just <em>correct</em>. It was already thinking one step ahead of me.</p><p>I sat back for a second. And that’s when it hit me.</p><p><strong>I wasn’t using AI to help me code. I was coding <em>with</em> AI.</strong></p><p>There’s a difference. A big one.</p><h3>The Tool I Thought I Knew</h3><p>For the first few months, AI in my workflow felt like autocomplete on steroids. Convenient. Fast. Sometimes annoying when it guessed wrong.</p><p>I’d write the logic, and it would fill in the boring parts — boilerplate, imports, the stuff nobody wants to type twice.</p><p>I thought that was the whole story.</p><p><strong>It wasn’t even the interesting part.</strong></p><h3>The Shift Nobody Warns You About</h3><p>Somewhere along the way, the conversation changed. I stopped writing code line by line and started <em>describing intent</em>:</p><ul><li>“Handle the empty state here.”</li><li>“This API call needs a retry with backoff.”</li><li>“Refactor this so it’s testable.”</li></ul><p>And it responded — not like a search engine spitting out a Stack Overflow answer, but like someone who’d actually read my code and understood <em>why</em> I wrote it that way.</p><p>I’d push back. It would adjust. I’d ask “why not do it this way instead?” — and get a real answer, with trade-offs.</p><p>That’s not autocomplete. That’s a conversation.</p><p><strong>Somewhere between the tenth prompt and the hundredth, my keyboard became a place where two minds met.</strong></p><h3>The Uncomfortable Question</h3><p>If I’m being honest, this realization didn’t feel entirely good at first.</p><p>For four years, my identity as a developer was tied to <em>how</em> I solved problems:</p><ul><li>The elegant function.</li><li>The clever fix at 2 AM.</li><li>The pride of writing something nobody else in the room could.</li></ul><p>So when AI started matching — sometimes beating — my first instinct, I had to ask myself something uncomfortable:</p><p><strong>If AI can write the code, what exactly am I bringing to the table?</strong></p><p>I didn’t like sitting with that question. But I sat with it anyway.</p><h3>What I Actually Bring</h3><p>Here’s what I found on the other side of that discomfort.</p><p>AI is brilliant at <em>how</em>:</p><ul><li>Write a function.</li><li>Fix a bug.</li><li>Suggest three ways to structure a widget tree.</li></ul><p>But it doesn’t know <strong>why</strong> this app exists:</p><ul><li>It doesn’t know our users are runners checking race results with shaky hands and bad signal.</li><li>It doesn’t know which shortcuts will bite us in six months.</li><li>It doesn’t know which corner is worth cutting and which one isn’t.</li></ul><p><strong>AI writes code. I write decisions.</strong></p><p>That’s the part that didn’t get automated. That’s the part that got <em>more</em> important.</p><p>My job quietly shifted — from typing every line, to knowing which lines matter. From being the one who writes the solution, to being the one who knows which problem is worth solving.</p><h3>Coding With, Not Coding Alone</h3><p>These days, my workflow looks less like “developer writes code” and more like:</p><ul><li><strong>Developer directs</strong> — I describe the problem and the constraints.</li><li><strong>AI drafts</strong> — it proposes a solution, fast.</li><li><strong>Developer decides</strong> — I keep what’s right, rewrite what isn’t, and own the outcome either way.</li></ul><p>I still review every line. I still say no — a lot. But I’m no longer alone in the room while I build.</p><p>And strangely, that’s made the work feel <em>more</em> creative, not less. I spend less time fighting syntax and more time thinking about what the app should actually feel like to use.</p><p><strong>The best developers I know aren’t the ones ignoring AI, or the ones blindly trusting it. They’re the ones who learned how to think alongside it.</strong></p><h3>The Realization That Stuck</h3><p>I still remember that Tuesday — not because anything dramatic happened, but because something quietly clicked.</p><p>I wasn’t using a smarter tool.</p><p>I was doing my job differently. Better, in some ways. More honestly, in others.</p><p><strong>And once you notice you’re coding <em>with</em> AI instead of just <em>using</em> it, you can’t really unsee it.</strong></p><p>You just start building differently — one honest decision at a time.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9ff254979161" width="1" height="1" alt=""><hr><p><a href="https://blog.stackademic.com/the-day-i-realized-i-was-coding-with-ai-not-just-using-it-9ff254979161">The Day I Realized I Was Coding With AI, Not Just Using It</a> was originally published in <a href="https://blog.stackademic.com/">Stackademic</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>