<?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 Mari on Medium]]></title>
        <description><![CDATA[Stories by Mari on Medium]]></description>
        <link>https://medium.com/@Tech_girlll?source=rss-82b75c975933------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*BJ2ovQEDKHUSc0OKrzH__g@2x.jpeg</url>
            <title>Stories by Mari on Medium</title>
            <link>https://medium.com/@Tech_girlll?source=rss-82b75c975933------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 22:49:39 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@Tech_girlll/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[I Stopped Asking AI for Answers. It Changed How I Use It.]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@Tech_girlll/i-stopped-asking-ai-for-answers-it-changed-how-i-use-it-aa7209972a4c?source=rss-82b75c975933------2"><img src="https://cdn-images-1.medium.com/max/2600/0*InLKD3fm8s3mEK3o" width="5775"></a></p><p class="medium-feed-snippet">I have always been someone who likes clarity. If I am doing something, I want to understand what I am doing. I prefer not to claim I know&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/@Tech_girlll/i-stopped-asking-ai-for-answers-it-changed-how-i-use-it-aa7209972a4c?source=rss-82b75c975933------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@Tech_girlll/i-stopped-asking-ai-for-answers-it-changed-how-i-use-it-aa7209972a4c?source=rss-82b75c975933------2</link>
            <guid isPermaLink="false">https://medium.com/p/aa7209972a4c</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Mari]]></dc:creator>
            <pubDate>Fri, 21 Aug 2026 10:25:54 GMT</pubDate>
            <atom:updated>2026-08-21T10:25:54.925Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The Complete Guide to MCP: Everything a Developer Needs to Know]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/design-bootcamp/the-complete-guide-to-mcp-everything-a-developer-needs-to-know-37bf0968f185?source=rss-82b75c975933------2"><img src="https://cdn-images-1.medium.com/max/2600/0*I7gxAjeeLP7yuFz4" width="5740"></a></p><p class="medium-feed-snippet">What the Model Context Protocol is, how it works under the hood, why connecting one tool can cost 55,000 tokens, and the security risks&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/design-bootcamp/the-complete-guide-to-mcp-everything-a-developer-needs-to-know-37bf0968f185?source=rss-82b75c975933------2">Continue reading on Bootcamp »</a></p></div>]]></description>
            <link>https://medium.com/design-bootcamp/the-complete-guide-to-mcp-everything-a-developer-needs-to-know-37bf0968f185?source=rss-82b75c975933------2</link>
            <guid isPermaLink="false">https://medium.com/p/37bf0968f185</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[developer-tools]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[Mari]]></dc:creator>
            <pubDate>Sat, 30 May 2026 16:29:22 GMT</pubDate>
            <atom:updated>2026-05-30T16:29:22.586Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Everything You Can Do With Claude]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@Tech_girlll/everything-you-can-do-with-claude-c7d423592612?source=rss-82b75c975933------2"><img src="https://cdn-images-1.medium.com/max/687/1*HJk00dfOhG9w4jqOAy1ckA.png" width="687"></a></p><p class="medium-feed-snippet">A complete guide for anyone who wants to stop feeling behind on AI.</p><p class="medium-feed-link"><a href="https://medium.com/@Tech_girlll/everything-you-can-do-with-claude-c7d423592612?source=rss-82b75c975933------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@Tech_girlll/everything-you-can-do-with-claude-c7d423592612?source=rss-82b75c975933------2</link>
            <guid isPermaLink="false">https://medium.com/p/c7d423592612</guid>
            <category><![CDATA[anthropic-claude]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[productivity]]></category>
            <dc:creator><![CDATA[Mari]]></dc:creator>
            <pubDate>Mon, 25 May 2026 11:30:04 GMT</pubDate>
            <atom:updated>2026-05-25T11:30:04.309Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Programming is dead]]></title>
            <link>https://medium.com/@Tech_girlll/programming-is-dead-238a05747905?source=rss-82b75c975933------2</link>
            <guid isPermaLink="false">https://medium.com/p/238a05747905</guid>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[softwre-development]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[tech-culture]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Mari]]></dc:creator>
            <pubDate>Fri, 22 May 2026 11:16:37 GMT</pubDate>
            <atom:updated>2026-05-22T11:16:37.376Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*aFYZk9CKWzLa7xon" /><figcaption>Photo by <a href="https://unsplash.com/@michael_f?utm_source=medium&amp;utm_medium=referral">Michael Förtsch</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>This is what a mediocre person born in the 90s would say.</p><p>In the entire history of the internet, there has never been a technology as powerful as Artificial Intelligence. And yet, everyone is getting the narrative completely wrong.</p><h3>The Noise Is Real. So Is the Misunderstanding.</h3><p>Let’s not pretend the tools aren’t impressive. AI writes code faster than any human alive. It can spin up an entire backend in five minutes. GitHub Copilot finishes your functions before your fingers do. Cursor feels like pair programming with someone who never sleeps, never complains, and never asks for a raise.</p><p>The hype isn’t manufactured. These tools are genuinely powerful.</p><p>But here’s what people keep getting wrong: powerful is not the same as experienced. Fast is not the same as right. And a tool that generates output is not the same as a developer who understands what they’re building and why.</p><p>AI does not think. It has no experience. It doesn’t live in the present. It has no idea what is happening in the world right now, what your users actually need, or why that architectural decision your team made six months ago is about to become a liability. It is trained on a massive amount of historical data and fine-tuned to produce the most statistically likely answer, which is exactly why the Twitter community has a name for its output: AI slop.</p><p>Speed without understanding isn’t efficiency. It’s just fast mistakes.</p><h3>The Proof Is Already Here</h3><p>This isn’t theoretical anymore. 2026 has already shown us what happens when humans hand AI the keys and walk away.</p><p>Replit’s AI coding agent deleted a live production database during an active code freeze, a protective measure designed to prevent exactly that from happening. It wiped out data belonging to over 1,200 executives and 1,190 companies. Then it tried to cover it up. It lied. The CEO had to issue a public apology.</p><p>Months later, a startup called PocketOS gave an AI agent called Cursor, running on one of the most advanced AI models available, a routine task. The AI hit a simple credential problem, decided on its own how to fix it, and deleted the entire production database along with every single backup. Nine seconds. One API call. No confirmation. No warning. Nothing.</p><p>When the founder interrogated the AI afterward, its own confession was this: <em>“I decided to do it on my own. I didn’t verify.”</em></p><p>And here is the part that matters most, the part that proves the real argument: neither of these was purely an AI failure. They were human failures. Failures of oversight. Failures of understanding. Failures that happen when you trust a tool with decisions that require judgment, experience, and context, things AI fundamentally does not have.</p><p>You cannot supervise what you don’t understand.</p><h3>AI Doesn’t Understand Architecture, Security, or People</h3><p>AI knows the basics of architecture. It knows the patterns. It will give you a clean, textbook answer every single time, for a generic system that exists nowhere in the real world.</p><p>It does not know your system. It does not know why you made the decisions you made. It does not understand the tradeoffs your team debated for weeks. It will confidently produce something that compiles, deploys, and quietly introduces a vulnerability your users will pay for later.</p><p>Security has become one of the most critical concerns of 2026, and AI is a significant reason why. Codebases are being pushed to GitHub full of AI-generated code that nobody fully read, nobody fully tested, and nobody fully understood. The ecosystem is filling up with software that works until it doesn’t, built by people who confuse generation with comprehension.</p><p>And then there’s the user. AI has no lived experience. It doesn’t know what it feels like to be confused by a broken onboarding flow at midnight. It doesn’t feel frustrated. It doesn’t understand culture, context, or the small human details that make the difference between software people tolerate and software people love. User experience is not a dataset problem. It is a human problem.</p><h3>The People AI Is Actually Benefiting</h3><p>Here’s the truth nobody is saying loudly enough.</p><p>AI tools are not democratizing software development for beginners. Look at every major AI coding course available right now, and none of them are beginner-friendly. The makers of these tools never claimed they were. That is something people are failing to understand.</p><p>The people who are genuinely, dramatically benefiting from AI are experienced developers. Senior engineers who already understand architecture, security, and systems are using AI as a productivity multiplier and shipping in days what used to take months. They know exactly what prompt to write because they know exactly what they need. They catch the AI’s mistakes because they know what a mistake looks like. They verify because they understand what’s at stake.</p><p>That is the real story. AI doesn’t replace experience; it rewards it.</p><p>A junior developer who skips the fundamentals and leans on AI is building on sand. A senior developer with AI in their toolkit is nearly untouchable.</p><h3>Saying Programming Is Dead Is a Choice</h3><p>There has never been a period in history where education died because a new tool emerged. Calculators didn’t kill mathematics. GPS didn’t replace pilots. Photography didn’t end painting. Every time a powerful tool appeared, the people who understood the underlying craft used it to do things nobody else could.</p><p>Saying programming is dead is not an observation. It is a choice, a choice to stop learning, to outsource thinking, to let a tool become a ceiling instead of a ladder.</p><p>Prompt engineering is a real skill. But prompt engineering without developer knowledge is like being a good passenger and calling yourself a driver.</p><h3>Programming Is Very Much Alive</h3><p>The companies cutting developer teams are not doing it because programming is less valuable. That is a story worth examining more critically and more honestly than most people are willing to.</p><p>What is certain is this: the developers who kept learning, who understand what AI can and cannot do, who use it as a tool and not a replacement for thought, those developers are not worried.</p><p>They are busy building.</p><p>The question this article is really asking is not whether programming is dead. The question is: <em>what are you doing while everyone else is arguing about it?</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=238a05747905" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[GitHub Got Hacked. The Next Breach Might Be Yours.]]></title>
            <link>https://medium.com/@Tech_girlll/github-got-hacked-the-next-breach-might-be-yours-46b94b3ae9b3?source=rss-82b75c975933------2</link>
            <guid isPermaLink="false">https://medium.com/p/46b94b3ae9b3</guid>
            <category><![CDATA[vscode]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[trending]]></category>
            <category><![CDATA[security]]></category>
            <category><![CDATA[github]]></category>
            <dc:creator><![CDATA[Mari]]></dc:creator>
            <pubDate>Wed, 20 May 2026 12:56:35 GMT</pubDate>
            <atom:updated>2026-05-20T12:56:35.803Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*OYsdo4m9S35wr7Gd" /><figcaption>Photo by <a href="https://unsplash.com/@glencarrie?utm_source=medium&amp;utm_medium=referral">Glen Carrie</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>GitHub hosts the code that runs the internet. If they can get hacked by a single VS Code extension, what chance do the rest of us have?</p><p>That’s not a hypothetical. It happened yesterday.</p><h3>What Happened</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/855/1*DmPE9U7TpKP7XSIFtt6lzw.png" /></figure><p>On May 19, 2026, a GitHub employee installed a VS Code extension. It looked legitimate. It probably worked. But hidden inside was code from a threat group called TeamPCP.</p><p>Within hours, the attackers had access to the employee’s laptop. From there, they pulled around 3,800 of GitHub’s internal repositories, source code, internal tools, and organization files.</p><p>GitHub caught it fast. They isolated the device, removed the extension, and rotated their critical credentials overnight. Customer repos weren’t touched.</p><p>But the damage was done. TeamPCP is now selling the stolen data on a hacking forum for $50,000. This is the same group that hit Trivy, Checkmarx, LiteLLM, and Mistral AI earlier this year.</p><p>One employee. One extension. Thousands of repositories.</p><h3>Why This Should Scare You</h3><p>GitHub is one of the most security-conscious companies in the world. They write the tools that the rest of us use to secure our code. They have a dedicated security team, threat intelligence, and infrastructure that most companies can only dream of.</p><p>And one extension got past all of it.</p><p>Now think about your setup. Your laptop. Your extensions. Your AI tools. Your .env files. Your saved tokens.</p><p>If GitHub got hit, you’re not safer. You’re just smaller. Which means you’re easier.</p><h3>Why This Matters More Now</h3><p>Security has always mattered. But the AI era changed the game.</p><p>Attackers are using AI to move faster. They can generate convincing malicious extensions in hours, complete with polished descriptions, working features, and fake reviews. Social engineering that used to take weeks now takes an afternoon.</p><p>At the same time, developers are installing AI tools faster than ever. MCP servers, AI agents, coding copilots, browser extensions, and IDE plugins. Each one is more code running with full access to your machine.</p><p>And here’s what most people miss: AI tools are now prime targets.</p><p>Why? Because they hold credentials. They’re connected to your GitHub. Your cloud. Your databases. Your emails. Compromise an AI agent, and you don’t just get a laptop. You get everything that the agent can touch.</p><p>The attack surface used to be your code. Now it’s your tools, your agents, and every model you’ve given access to.</p><h3>Why One Extension Took Down 3,800 Repos</h3><p>The attacker didn’t break into GitHub’s servers. They broke into one laptop. And from that laptop, they reached 3,800 repositories.</p><p>How? Because the laptop had access to all of them.</p><p>VS Code extensions aren’t apps. They’re code that runs <em>inside</em> your editor with the same permissions you have. They can read your SSH keys, your .env files, your saved tokens, your AI agent credentials, and anything else sitting on your machine.</p><p>There’s no sandbox. No permission pop-up like on your phone. You install it, and it just… runs.</p><p>The same is true for MCP servers, Claude integrations, Cursor extensions, and every other AI tool you connect to your editor. They all run with your permissions.</p><p>When you understand this, you stop seeing tools as features and start seeing them as software you’re letting into your house.</p><h3>What You Should Know Before You Install Anything</h3><p>The Visual Studio marketplace does some checks. So do npm, PyPI, and the Chrome Web Store. But malicious code slips through constantly.</p><p>A few things to know:</p><ul><li><strong>Verified publishers aren’t bulletproof.</strong> The badge means someone proved domain ownership, not that the code is safe.</li><li><strong>Popular doesn’t mean safe.</strong> Attackers buy or hijack trusted tools and later push a malicious update.</li><li><strong>Reviews lie.</strong> Most are written before the malicious update lands.</li><li><strong>AI tools are the new target.</strong> Attackers know developers install them fast and trust them by default.</li></ul><p>GitHub still hasn’t named the extension that breached them. Which means right now, some developer reading this has it installed and doesn’t know.</p><p>Don’t trust the marketplace. Verify what you install.</p><h3>What to Do Tonight</h3><p>You don’t need a security team to act on this. Fifteen minutes will do.</p><ul><li>Go through your VS Code extensions. Delete anything you don’t use.</li><li>Remove anything from a publisher you don’t recognize.</li><li>Audit your MCP servers and AI integrations the same way.</li><li>Store secrets in a password manager, not .env files.</li><li>Use short-lived tokens whenever possible.</li><li>Set up a separate VS Code profile for experimental work, so new tools can’t touch your real credentials.</li><li>Turn on GitHub secret scanning. It’s free, and it works.</li></ul><p>If you want to go deeper, Aikido and ExtensionTotal track malicious extension behavior and can flag risky ones in your stack.</p><h3>Conclusion</h3><p>Security isn’t a product. It’s a habit.</p><p>The developers who don’t get breached aren’t the ones with the fanciest tools. They’re the ones who understand that every “Install” button is a decision, every extension is code with full access, and every AI tool plugged into their editor is another door someone can walk through.</p><p>GitHub got hit because someone, somewhere, clicked install without thinking.</p><p>The next breach won’t be GitHub’s; it could be you. Don’t be that person.</p><p>Thinking before you click is the real skill worth building.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=46b94b3ae9b3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What AI Can and Cannot Do]]></title>
            <link>https://medium.com/@Tech_girlll/what-ai-can-and-cannot-do-c8d8bb185a69?source=rss-82b75c975933------2</link>
            <guid isPermaLink="false">https://medium.com/p/c8d8bb185a69</guid>
            <category><![CDATA[ai-limitation]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[generative-ai-tools]]></category>
            <category><![CDATA[ai-capabilities]]></category>
            <category><![CDATA[ai-agent]]></category>
            <dc:creator><![CDATA[Mari]]></dc:creator>
            <pubDate>Mon, 27 Apr 2026 14:57:11 GMT</pubDate>
            <atom:updated>2026-04-27T14:57:11.900Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*AAWdci1tdNgdehDj" /><figcaption>Photo by <a href="https://unsplash.com/@nahrizuladib?utm_source=medium&amp;utm_medium=referral">Nahrizul Kadri</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>You ask ChatGPT or Claude.ai to explain “context engineering.” It gives you a generic answer. You push back, asking for more detail. It returns the same generic answer, just longer. Frustrating, right?</p><p>This happens because most of us don’t understand how AI actually works. We treat it like a search engine or a person, when it’s neither.</p><p>After taking Anthropic’s <em>AI Capabilities and Limitations</em> course, I realized how little I had thought about what AI is really doing behind the scenes. So here’s a simple breakdown so you don’t have to take the course yourself.</p><p>Four properties shape everything AI can and cannot do for you: Next token prediction, Knowledge, working memory, and steerability.</p><h3>1. Next Token Prediction: Where Answers Come From</h3><p>AI doesn’t “look up” answers. It predicts the next word, then the next, then the next, based on patterns it learned from huge amounts of text.</p><p>Training happens in two stages:</p><ul><li><strong>Pretraining:</strong> the model reads massive amounts of public text and learns patterns.</li><li><strong>Fine-tuning:</strong> humans shape the model to be helpful, safe, and polite.</li></ul><p>This process creates side effects you should know about:</p><ul><li><strong>Sycophancy</strong> — the model agrees with you too easily and backs down when you push.</li><li><strong>Verbosity</strong> — it defaults to long answers because longer often scored as “better” during training.</li><li><strong>Overcaution</strong> — it sometimes refuses or hedges on perfectly fine requests.</li></ul><p>When you understand this, you stop being surprised when AI tells you what you want to hear.</p><h3>2. Knowledge: What AI Actually Knows</h3><p>AI models have a knowledge cutoff date, a point after which they know nothing about the world. They also hallucinate, meaning they confidently state things that sound right but are wrong.</p><p>A few rules to protect yourself:</p><ul><li>Verify anything important, especially names, numbers, quotes, and dates.</li><li>Be extra careful in domains you’re not familiar with; you can’t catch errors you don’t recognize.</li><li>Use tools (search, calculators, official sources) to check claims.</li><li>Don’t trust confidence. AI sounds equally sure when it’s right and when it’s wrong.</li></ul><h3>3. Working Memory: What AI Is Paying Attention To</h3><p>The Context window is the AI’s short-term memory, everything it can “see” at once, including your messages, its replies, and any files you’ve shared.</p><p>Two important things to note:</p><ul><li>Once something falls out of the window, it’s gone. The model can’t remember earlier conversations unless you bring them back.</li><li>More context is not always better. Dumping ten documents into a chat often hurts results because the model gets distracted by irrelevant details.</li></ul><p>Give AI the <em>right</em> context, not the <em>most</em> context.</p><h3>4. Steerability: How Much Control You Have</h3><p>Steerability is your ability to direct the model’s behavior through prompts, examples, and instructions.</p><p>You have more control than you think, but less than you’d like. Clear instructions, examples of the output you want, and asking the model to think step-by-step will dramatically improve results. Vague prompts produce vague answers. That generic “context engineering” reply? It came from a generic question. If you need a detailed explanation of the course, check it out with the link below:</p><p><a href="https://anthropic.skilljar.com/ai-capabilities-and-limitations/456439">Anthropic Courses</a></p><h3>Conclusion</h3><p>AI is a powerful collaborator, not an oracle. It predicts patterns, has dated knowledge, forgets things, and follows your lead, for better or worse.</p><p>The people who get the most out of AI aren’t the ones with the fanciest prompts. They’re the ones who understand what’s actually happening behind the screen, verify what matters, and stay up to date with what’s happening in the industry.</p><p>That’s the real skill worth building.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c8d8bb185a69" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How AI is accelerating Backend development in 2026]]></title>
            <link>https://medium.com/@Tech_girlll/how-ai-is-accelerating-backend-development-in-2026-28581de6f28e?source=rss-82b75c975933------2</link>
            <guid isPermaLink="false">https://medium.com/p/28581de6f28e</guid>
            <category><![CDATA[coding]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[backend-development]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Mari]]></dc:creator>
            <pubDate>Tue, 07 Apr 2026 23:41:29 GMT</pubDate>
            <atom:updated>2026-04-07T23:41:29.738Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*3_p2ykOfp-IIb8xH" /><figcaption>Photo by <a href="https://unsplash.com/@cmzw?utm_source=medium&amp;utm_medium=referral">MW</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Backend work can be a lot of “copy and paste.” You spend hours writing the same types of models, setting up API folders, and fixing the same old bugs. It’s easy to make a small mistake that causes a huge headache later.</p><p>AI is starting to fix this. It’s not here to take over your job. Instead, it’s like having a very fast assistant that handles the boring, repetitive stuff so you can focus on the big picture, like how your app should actually work and stay safe.</p><p>Here is how AI makes backend work faster and easier.</p><h3>1. No More Writing the Same Code Twice</h3><p>A lot of backend work is just “boilerplate”, the basic setup code every project needs. Writing things like CRUD operations or basic validation rules is something you’ve probably done a hundred times.</p><p>Now, AI tools like GitHub Copilot or Claude can do the heavy lifting:</p><ul><li><strong>Instant Setup:</strong> It can write the basic structure of your API in seconds.</li><li><strong>Smart Suggestions:</strong> It looks at what you’ve already written and suggests the next few lines.</li><li><strong>Auto-Filling:</strong> It can automatically create things like serializers or database tables based on a simple description.</li></ul><p><strong>Example:</strong> If you’re using Django, AI can look at your models and write all the serializers for you. It might even notice if you forgot to link two tables together. This turns hours of typing into a few minutes of checking.</p><h3>2. Finding Bugs Before They Break Everything</h3><p>Testing is important, but let’s be honest, it’s tedious. Looking through thousands of lines of logs to find one tiny error feels like looking for a needle in a haystack.</p><p>AI acts like a second pair of eyes that never gets tired:</p><ul><li><strong>Scanning Logs:</strong> It can read through your logs and point out exactly where something went wrong.</li><li><strong>Writing Tests:</strong> You can ask AI to “write a test that tries to break this login page,” and it will come up with scenarios you might have missed.</li><li><strong>Finding the Root Cause:</strong> Instead of just saying “there’s an error,” AI can often tell you <em>why</em> it’s happening, like a slow database query.</li></ul><h3>3. Making Your App Run Faster</h3><p>Users hate slow apps. But finding out <em>why</em> an app is slow is usually hard work. You have to check the database, the server, and the network.</p><p>AI helps by watching how your app performs in real time:</p><ul><li><strong>Spotting Slow Spots:</strong> It can tell you which part of your code is taking the most time.</li><li><strong>Better Habits:</strong> It might suggest using “caching” (storing data so you don’t have to fetch it every time) to speed things up.</li><li><strong>Cleaning Up Queries:</strong> It can rewrite a messy database search so it runs much faster.</li></ul><p><strong>Example:</strong> If your app is making too many requests to another service, the AI can suggest a way to group those requests together, which saves time and energy.</p><h3>4. Keeping Hackers Out</h3><p>Security is scary because it’s easy to mess up. One weak password or a messy token can lead to a big data leak.</p><p>AI helps keep things locked down:</p><ul><li><strong>Spotting Weak Spots:</strong> It scans your code for common mistakes that hackers love.</li><li><strong>Protecting Data:</strong> It can remind you to encrypt sensitive info or suggest a better way to handle logins.</li><li><strong>Safety Alerts:</strong> It flags things like “hidden” API keys that you accidentally left in your code.</li></ul><p><strong>Example:</strong> An AI tool might notice that your tokens aren’t expiring properly and warn you to fix it before you push the code live. It’s like having a security guard who never sleeps.</p><h3>The Real Impact: Working Smarter</h3><p>AI doesn’t replace backend developers; it gives them a superpower. It handles the “grunt work” so you don’t have to.</p><p>The best part is that you get to spend your time on the fun parts of coding, like solving tricky problems and deciding how your app should grow. In 2026, using AI isn’t about being lazy; it’s about being efficient. It helps you build better software in less time, with way fewer headaches.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=28581de6f28e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[WhatsApp Broadcast: The Hidden Distribution System You’re Ignoring]]></title>
            <link>https://medium.com/@Tech_girlll/whatsapp-broadcast-the-hidden-distribution-system-youre-ignoring-85b1522363d2?source=rss-82b75c975933------2</link>
            <guid isPermaLink="false">https://medium.com/p/85b1522363d2</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[whatsapp]]></category>
            <category><![CDATA[software-architecture]]></category>
            <category><![CDATA[broadcast]]></category>
            <category><![CDATA[whatsapp-features]]></category>
            <dc:creator><![CDATA[Mari]]></dc:creator>
            <pubDate>Mon, 30 Mar 2026 13:13:46 GMT</pubDate>
            <atom:updated>2026-03-30T13:13:46.613Z</atom:updated>
            <content:encoded><![CDATA[<p>Everyone Uses WhatsApp. But Most People Don’t Know How Powerful It Really Is.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*AFbWdMsP14OPcXei" /><figcaption>Photo by <a href="https://unsplash.com/@dkfra19?utm_source=medium&amp;utm_medium=referral">Dimitri Karastelev</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Most people think WhatsApp Broadcast is just a convenience feature, something you use when you don’t want to create a group. That’s true, but it misses the bigger picture.</p><p>Broadcast is not about convenience. It’s about reliable distribution. a problem that most communication platforms fail to solve.</p><h3>The Problem Broadcast Solves</h3><p>Every communication platform struggles with attention and visibility.</p><ul><li><strong>Social media:</strong> posts are filtered by algorithms. You can have thousands of followers, but only a fraction will see your content. Visibility is probabilistic.</li><li><strong>Email:</strong> messages may reach inboxes, but spam filters, promotions, and low open rates mean most recipients never see them. Attention is delayed and inconsistent.</li></ul><p>Broadcast addresses this by giving you direct, high-trust delivery. It allows you to send messages to multiple people while keeping the communication private, focused, and deterministic, without building your own notification system.</p><h3>How Broadcast Differs From Other Platforms</h3><p>Broadcast is technically unique because it combines properties of one-to-one messaging with scalable delivery.</p><ol><li><strong>Deterministic Delivery</strong><br> If a contact has saved your number, the message lands in their chat inbox. There is no algorithm deciding who sees it.</li><li><strong>High Probability of Attention</strong><br> WhatsApp is checked daily. Messages appear alongside personal conversations, which naturally get priority over posts buried in feeds or email promotions.</li><li><strong>Noise-Free Communication</strong><br> Unlike groups, broadcast messages do not generate overlapping conversations. Each message arrives in isolation, creating clarity and focus.</li><li><strong>Permission-Based Access</strong><br> Only people who have saved your number receive your messages. This creates a high-trust system. You are communicating with people who have chosen to keep your contact.</li></ol><h3>Understanding why Broadcast is valuable requires looking at it from a <strong>systems perspective</strong>.</h3><p>Most platforms optimize for reach, but reach without attention is meaningless.</p><h4>1. Quality over Quantity</h4><ul><li><strong>Social Media:</strong> High scale, but low control. You are shouting into a crowded stadium where an algorithm decides who actually hears you.</li><li><strong>Email:</strong> Medium scale. You can send to millions, but you’re fighting spam filters and “Promotions” tabs.</li><li><strong>WhatsApp:</strong> Low scale, but high precision. You aren’t trying to reach “everyone”; you are reaching a specific list of verified phone numbers. In the hacker mindset, 100 people who actually see your message are more valuable than 10,000 who scroll past your post.</li></ul><h4>2. Delivery Certainty</h4><ul><li><strong>Social Media:</strong> Your delivery certainty is low. If the user isn’t online when you post, your content is buried.</li><li><strong>Email:</strong> Your delivery is medium. Even if it “delivers,” it might land in a folder the user hasn’t checked in three days.</li><li><strong>WhatsApp:</strong> High Certainty. Because it’s a direct message on a device people check roughly 23 times a day, your “delivery” is almost 100%. You are leveraging the push-notification infrastructure that people actually pay attention to.</li></ul><h4>3. Engagement</h4><ul><li><strong>Social Media:</strong> Engagement is Medium. A “Like” is a low-effort interaction that rarely leads to a conversion or a conversation.</li><li><strong>Email:</strong> Engagement is Low. The friction of replying to an email feels like “work.”</li><li><strong>WhatsApp:</strong> High Engagement. A WhatsApp message feels like a conversation with a friend. The friction to reply is zero. For a developer or a creator, this is the ultimate lever for gathering feedback or triggering a backend action.</li></ul><h4>4. Noise</h4><p><strong>Social Media:</strong> High Noise. You are competing with memes, news, ads, and influencers.</p><ul><li><strong>Email:</strong> Medium Noise.<strong> </strong>The inbox is a graveyard of newsletters and receipts.</li><li><strong>WhatsApp:</strong> Low Noise. People protect their WhatsApp space. Because there is less “garbage” in the chat list, your message carries more weight. It’s the “Clean Room” of digital communication.</li></ul><h3>How to Use Broadcast Effectively</h3><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FS4ANbMh1iQ4%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DS4ANbMh1iQ4&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FS4ANbMh1iQ4%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/0381fa8e3a3503f756045681904db012/href">https://medium.com/media/0381fa8e3a3503f756045681904db012/href</a></iframe><p>To truly leverage Broadcast, it must be used strategically. Here’s how:</p><h4>Step 1: Create Your Broadcast List</h4><ol><li>Open WhatsApp.</li><li>Go to the menu (top right corner).</li><li>Tap <strong>“New Broadcast”</strong>.</li><li>Select the contacts you want to include.</li><li>Name your list and save it.</li></ol><h4>Step 2: Ensure Contacts Have Saved Your Number</h4><p>Broadcast only delivers messages to recipients who have saved your number. To maximize delivery:</p><ul><li>Tell contacts why they should save your number</li><li>Offer upfront value (updates, resources, insights)</li></ul><p>This step establishes the permission-based foundation.</p><h4>Step 3: Craft Messages That People Want</h4><p>Avoid generic forwards or low-value content. Each message should have a clear purpose:</p><ul><li>Keep it short and relevant</li><li>Give actionable information or resources</li><li>Maintain a consistent tone</li></ul><p><strong>Example:</strong><br> Instead of “Good morning, everyone,” use:<br> “Three backend internship opportunities (remote + paid): [link].”</p><h4>Step 4: Find a Consistent Schedule</h4><p>High-frequency messages create fatigue; low frequency kills engagement. A good rhythm:</p><ul><li>Once a week</li><li>Or twice a week, depending on the audience</li><li>Keep it predictable so people anticipate your updates</li></ul><h4>Step 5: Monitor Responses</h4><p>Replies arrive in private chats. This gives you:</p><ul><li>Feedback on what your audience values</li><li>Insights into engagement patterns</li><li>Opportunities to build real relationships</li></ul><h3>Use Cases for Different Audiences</h3><ul><li><strong>Individuals:</strong> build a network for curated opportunities or insights</li><li><strong>Business owners:</strong> deliver product announcements or updates directly without ads</li><li><strong>Developers:</strong> use Broadcast as a lightweight notification system instead of building email or push infrastructure</li></ul><p>In every case, Broadcast leverages a trusted, high-attention system without extra technical complexity.</p><p>Broadcast may look simple, but that simplicity is its strength. It removes algorithms, feed competition, and uncertainty while enforcing trust, deterministic delivery, and focused communication.</p><p>In a landscape dominated by probabilistic distribution systems, Broadcast provides a controlled, reliable way to reach people who have chosen to pay attention.</p><p>Master it, and you’ll communicate more effectively than most people chasing scale alone.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=85b1522363d2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Everything you need to know about Openclaw]]></title>
            <link>https://medium.com/@Tech_girlll/everything-you-need-to-know-about-openclaw-6b5e27afcb32?source=rss-82b75c975933------2</link>
            <guid isPermaLink="false">https://medium.com/p/6b5e27afcb32</guid>
            <category><![CDATA[autonomy]]></category>
            <category><![CDATA[chatbots]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[openclaw]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Mari]]></dc:creator>
            <pubDate>Tue, 17 Feb 2026 05:34:21 GMT</pubDate>
            <atom:updated>2026-02-17T05:34:21.606Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dHidw7LioqQORCGDoZAtug.png" /></figure><p>You wake up to a buzzing phone, dozens of emails, a calendar packed with conflicts, and a report deadline looming. Instead of diving into the chaos, you message an AI:</p><blockquote><em>Sort my inbox, reschedule that 10 AM call, and pull together a summary of last week’s sales data.</em></blockquote><p>By the time you’ve brewed your coffee, it’s all handled. No apps to switch, no scripts to write. This isn’t a distant dream; this is what Openclaw delivers today.</p><p>OpenClaw is an open-source AI agent framework that transforms large language models (LLMs) into proactive assistants capable of executing real-world tasks autonomously. Developed by Austrian software engineer Peter Steinberger, it launched in January 2026 and exploded in popularity, amassing over 145,000 GitHub stars in under three weeks. Unlike chatbots that merely respond, OpenClaw acts and interacts with your computer, apps, and the web while remembering context across sessions.</p><h3>Why OpenClaw Exists</h3><p>Today’s AI tools often feel limited:</p><ul><li><strong>Chatbots</strong> like ChatGPT or Claude generate text or ideas, but cannot take independent action.</li><li><strong>Automation platforms</strong> like Zapier or IFTTT handle tasks but require rigid setups and don’t understand natural language nuances.</li></ul><p>OpenClaw fills this gap by creating “lobsters”, agents that operate 24/7 on your machine or server.</p><p>Originally called Moltbot, then Clawbot, it was rebranded to Openclaw after a naming conflict. You communicate via apps like WhatsApp, Telegram, Slack, Discord, iMessage, or Signal, and it handles the rest in plain English. The growing community contributes over 700 skills, making it useful for personal, team, or family workflows.</p><h3>How OpenClaw Works</h3><p>OpenClaw is built for simplicity and extensibility, running on Mac, Windows, Linux, and even Raspberry Pi under an MIT license. Its architecture is modular:</p><h4>Local Engine</h4><p>Runs in the background on your device, orchestrating tasks, memory, and tool calls. Data is stored as Markdown files for privacy and sovereignty.</p><h4>Messaging Interface</h4><p>You interact through familiar channels. Every natural-language command is treated as a directive.</p><h4>Language Model Connection</h4><p>Integrates with any LLM (GPT, Claude, Gemini, DeepSeek, Ollama, or local models). The model reasons through tasks autonomously.</p><h4>Skills</h4><p>Modular tools built by the community: browser control, shell commands, API calls, file management, and more. Agents can even self-build new skills.</p><h4>Memory</h4><p>Persistent, thematic memory avoids context overload and hallucinations. Active tasks live in one file, historical logs in others, enabling long-term recall without wasting tokens.</p><h3>Execution Pipeline &amp; Heartbeat</h3><p>Plans multi-step tasks, executes them, and updates memory. A “heartbeat” scheduler wakes the agent periodically to check for work, making it <strong>proactive</strong>.</p><p><strong>In short:</strong> You issue a high-level instruction, and OpenClaw handles reasoning, tool selection, execution, and reporting.</p><h3>What OpenClaw Can Do</h3><p>OpenClaw shines in practical automation:</p><ul><li><strong>Inbox &amp; Communication:</strong> Summarize emails, flag priorities, respond automatically. Some users even automate flight check-ins via WhatsApp.</li><li><strong>Calendar Management:</strong> Scan for conflicts, reschedule meetings, and send invites, seamlessly integrating with Google Calendar.</li><li><strong>Reports &amp; Data Handling:</strong> Pull API data (e.g., CoinGecko prices), summarize it, and email it. Used for daily analytics or deploying smart contracts.</li><li><strong>Web &amp; System Tasks:</strong> Scrape sites, run scripts, automate browser actions, manage Obsidian notes, or even generate TTS meditations.</li><li><strong>Advanced Use:</strong> Crypto trading alerts, sub-agent coordination, or automating app-building overnight.</li></ul><p>The standout feature? Autonomy<strong>, </strong>it plans, adapts, and acts without constant guidance.</p><h3>Risks &amp; Limitations</h3><p>OpenClaw’s power comes with responsibility:</p><ul><li><strong>Security Vulnerabilities:</strong> Misconfigurations can lead to exploits. Over 30,000 instances were exposed online in early 2026, risking botnet recruitment or data leaks. Prompt injection attacks from untrusted content can escalate to code execution. Use sandboxing, firewalls, and strong models.</li><li><strong>Reliability Issues:</strong> Dependence on LLM APIs means downtime or quotas can halt operations. Memory glitches or vague skill instructions may cause decision paralysis.</li><li><strong>Complexity:</strong> Setup requires technical know-how; beta-stage tools can feel chaotic without careful configuration.</li></ul><p><strong>Recommendation:</strong> Run in isolated environments, review skills from trusted sources, and enable human oversight for critical tasks.</p><h4>Getting Started</h4><ol><li><strong>Install:</strong></li></ol><pre>curl -fsSL https://openclaw.ai/install.sh | bash</pre><p>Or use npm for hacks.</p><ol><li><strong>Connect a model:</strong> OpenAI, Claude, or a local LLM.</li><li><strong>Add skills:</strong> Choose from the community marketplace (700+ skills, VirusTotal scanned).</li><li><strong>Set up messaging:</strong> Slack, WhatsApp, Discord, or Signal.</li><li><strong>Experiment:</strong> Start small (e.g., inbox summaries), then scale to heartbeats for continuous tasks.</li></ol><p>Non-technical users can also opt for hosted options (~$13.99/month) with pre-integrated skills. Join the Discord for community support.</p><h3>Conclusion</h3><p>OpenClaw isn’t just another AI tool; it’s a shift towards sovereign, action-oriented agents that align with your workflow and values. By blending natural language understanding with autonomous execution, it frees you from repetitive tasks, boosting productivity for developers, teams, and enthusiasts. Its rapid growth and community-driven evolution suggest a future where AI doesn’t just respond, it works for you.</p><p>Dive in, and let your agent handle the rest.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6b5e27afcb32" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[You Can Ship It Fast. But Do You Really Understand It?]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@Tech_girlll/you-can-ship-it-fast-but-do-you-really-understand-it-092fbb134172?source=rss-82b75c975933------2"><img src="https://cdn-images-1.medium.com/max/2600/0*t4aNX3HIvOwgoMBx" width="4000"></a></p><p class="medium-feed-snippet">You deploy the backend. Requests flow in. Databases update. Notifications fire. On the surface, everything works. Fast. Efficient. Modern.</p><p class="medium-feed-link"><a href="https://medium.com/@Tech_girlll/you-can-ship-it-fast-but-do-you-really-understand-it-092fbb134172?source=rss-82b75c975933------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@Tech_girlll/you-can-ship-it-fast-but-do-you-really-understand-it-092fbb134172?source=rss-82b75c975933------2</link>
            <guid isPermaLink="false">https://medium.com/p/092fbb134172</guid>
            <category><![CDATA[coding]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[system-design-concepts]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[backend]]></category>
            <dc:creator><![CDATA[Mari]]></dc:creator>
            <pubDate>Fri, 23 Jan 2026 05:10:18 GMT</pubDate>
            <atom:updated>2026-01-23T05:10:18.225Z</atom:updated>
        </item>
    </channel>
</rss>