<?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 Stephanie Garmal on Medium]]></title>
        <description><![CDATA[Stories by Stephanie Garmal on Medium]]></description>
        <link>https://medium.com/@stephaniegarmal?source=rss-279b17d719c5------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*i8a-ehgmDGHsTt9hCotiAw.jpeg</url>
            <title>Stories by Stephanie Garmal on Medium</title>
            <link>https://medium.com/@stephaniegarmal?source=rss-279b17d719c5------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 12:31:08 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@stephaniegarmal/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[Should You Still Learn to Code in 2026? Yes, But Not for the Reason You Think]]></title>
            <link>https://medium.com/@stephaniegarmal/should-you-still-learn-to-code-in-2026-yes-but-not-for-the-reason-you-think-c9bd964a91b7?source=rss-279b17d719c5------2</link>
            <guid isPermaLink="false">https://medium.com/p/c9bd964a91b7</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[careers]]></category>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[Stephanie Garmal]]></dc:creator>
            <pubDate>Tue, 17 Mar 2026 14:49:07 GMT</pubDate>
            <atom:updated>2026-03-17T14:49:07.909Z</atom:updated>
            <content:encoded><![CDATA[<p><em>AI has made it easy to produce something that looks finished. That’s a much harder problem than a bug. Why technical foundations still matter, even if you never write production code.</em></p><p>A few months ago, someone showed me a project they had built entirely with AI. No coding background, no technical knowledge. They were proud of it, and honestly, rightfully so. It looked clean, it worked, and they had gotten there just by describing what they wanted.</p><p>So I ran a security audit and a code quality review on it.</p><p>The application was still running, and nothing was catastrophically broken, but there were enough gaps that if this had been a larger system, or something with real user data at stake, we would have been having a very different conversation. The kind that happens after something goes wrong.</p><h3>AI can get you far, just not all the way</h3><p>AI has made it genuinely easy to produce something that looks finished, so easy that a lot of people have started confusing the appearance of quality with actual quality, and that’s a much harder problem to fix than a bug. The person who built that project saw it running and thought they didn’t need to understand what was underneath, but what they actually proved is that AI can get you surprisingly far on a small, isolated thing and skip the parts that matter when no one is checking.</p><h3>What I keep telling business and product teams</h3><p>If your team has someone who understands AI but doesn’t understand development, that person cannot make good decisions on their own. They’ll rely on whatever the model suggests, and sometimes that’s fine, but sometimes it will have security gaps, no code standards, and technical debt baked in from day one. For a small project, you might survive that, but for a real system, you won’t.</p><p>I’ve had this conversation more than once with leadership teams I’ve worked with. You need people who can speak both languages, not necessarily someone who writes code all day, but someone who understands enough to evaluate what AI produced and know when to push back on what the model decided for you without asking.</p><h3>The real skill is knowing when to push back</h3><p>There’s a tension that even experienced developers feel right now, and it’s more complicated than most people admit. But what sits in the middle of it is exactly this: enough of a foundation to ask the right questions, catch the right gaps, and notice when AI is making a decision you didn’t realize it was making.</p><p>Learning to code in 2026 is not about competing with AI. It’s about being qualified to work with it.</p><p><em>This is the fifth article in a series about what AI adoption really looks like from the inside.</em></p><p><a href="https://medium.com/@stephaniegarmal/from-7-degrees-to-ai-a-nonlinear-career-path-f49f7f06d84b">From 7 Degrees to AI</a> · <a href="https://medium.com/@stephaniegarmal/ai-in-the-real-world-why-legacy-codebases-are-the-true-test-of-ai-adoption-4942f9ec1f1d">AI in Legacy Codebases</a> · <a href="https://medium.com/@stephaniegarmal/ai-wont-fix-your-process-388b37ad88e2">AI Won’t Fix Your Process</a> · <a href="https://medium.com/@stephaniegarmal/the-people-pushing-back-on-ai-arent-the-ones-you-should-worry-about-9e41f0c90520">The People Pushing Back on AI</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c9bd964a91b7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The People Pushing Back on AI Aren’t the Ones You Should Worry About]]></title>
            <link>https://medium.com/@stephaniegarmal/the-people-pushing-back-on-ai-arent-the-ones-you-should-worry-about-9e41f0c90520?source=rss-279b17d719c5------2</link>
            <guid isPermaLink="false">https://medium.com/p/9e41f0c90520</guid>
            <category><![CDATA[ai-adoption]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[tech-leadership]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Stephanie Garmal]]></dc:creator>
            <pubDate>Tue, 10 Mar 2026 14:53:09 GMT</pubDate>
            <atom:updated>2026-03-10T21:03:49.069Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Everyone focuses on getting the skeptics on board, but the real risk is the people who already think AI can do everything. The strongest developers are usually the most resistant, and that’s exactly right.</em></p><p>I’ve been training teams on AI across different levels of technical experience, and the pattern I keep seeing is not the one most people expect. The biggest challenge isn’t convincing the skeptics to try AI; it’s managing the people who are already convinced AI is going to do the work for them.</p><h3>The skeptics have a point</h3><p>The strongest developers I’ve worked with are usually the most resistant to AI. They tried it, it didn’t make the decisions they would have made, and they moved on. They’re not being stubborn; they’re being precise, because these are people who spent years getting really good at their craft, and when AI gives them something that doesn’t meet their standard, they lose patience with it fast. But those are exactly the people you want using AI, because they have the judgment to catch when it gets something wrong.</p><h3>The enthusiasm is the problem</h3><p>The people most excited about AI tend to be the ones with the least technical depth. They see AI generating code, and they think it’s going to do the work for them, and in some cases, it looks like it does, especially when the team doesn’t have a strong development process to begin with. Nobody is checking whether the output is actually good. They don’t review patterns, they don’t validate the architecture, and when AI hallucinates or makes a decision that breaks something, they don’t have the experience to notice. I’ve seen AI delete databases and generate code that quietly breaks things downstream, and the enthusiasts don’t question any of it because the output looks right.</p><h3>Training isn’t the real question</h3><p>Everyone wants to provide the entire company with AI tools and train everyone equally, and I understand the impulse. But not everyone should be using AI the same way. Someone who writes good specs and understands what they’re asking for will get completely different results than someone who doesn’t check what AI gives back. The people who get the most out of AI are the ones who are already curious and disciplined enough to research it on their own, who experiment carefully and actually want to understand how it works before they rely on it.</p><p>The judgment behind the tool matters more than the tool itself, and no amount of training will replace that judgment.</p><p><em>This is the fourth article in a series about what AI adoption really looks like from the inside.</em></p><p><a href="https://medium.com/@stephaniegarmal/from-7-degrees-to-ai-a-nonlinear-career-path-f49f7f06d84b"><em>From 7 Degrees to AI</em></a><em> · </em><a href="https://medium.com/@stephaniegarmal/ai-in-the-real-world-why-legacy-codebases-are-the-true-test-of-ai-adoption-4942f9ec1f1d"><em>AI in Legacy Codebases</em></a><em>· </em><a href="https://medium.com/@stephaniegarmal/ai-wont-fix-your-process-388b37ad88e2"><em>AI Won’t Fix Your Process</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9e41f0c90520" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Won’t Fix Your Process]]></title>
            <link>https://medium.com/@stephaniegarmal/ai-wont-fix-your-process-388b37ad88e2?source=rss-279b17d719c5------2</link>
            <guid isPermaLink="false">https://medium.com/p/388b37ad88e2</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[tech-leadership]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai-adoption]]></category>
            <dc:creator><![CDATA[Stephanie Garmal]]></dc:creator>
            <pubDate>Tue, 03 Mar 2026 14:54:07 GMT</pubDate>
            <atom:updated>2026-03-03T14:54:07.641Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Everyone’s optimizing their AI prompts, but they should be optimizing their process.</em></p><p>I’ve been bringing AI into development workflows across multiple teams with different codebases and levels of maturity. And after enough projects, a pattern became obvious: the quality of what AI gives you has almost nothing to do with the tool, it has everything to do with the process your team already has in place.</p><p>When a team has clear specs, consistent patterns, and real documentation, AI just works. You give it context, and it follows the structure, and what comes back requires minimal rework. The foundation does the heavy lifting.</p><p>When a team has none of that, AI still works, just in the wrong direction. It generates code that matches whatever inconsistency is already in the codebase, it passes tests that don’t actually test anything, and it moves fast while nobody notices it’s just continuing the mess the team already had.</p><h3>The problem is upstream</h3><p>The conversation right now is about which AI tool is best, which model writes better code, and how to write the perfect prompt, but none of that matters if what you’re feeding the AI is chaos. If you have no specs, AI is guessing what you want, and if you have no coding standards, it’s making its own decisions. The team thinks they’re moving faster because the output looks right, but when something breaks in production, and there’s no real development process behind it, no one even understands why.</p><h3>Your process is the real prompt</h3><p>The best thing you can do before adopting AI in your development workflow isn’t choosing a tool. It’s making sure your team has the basics in place. Clear specifications, consistent patterns, and a review process that actually holds code to a standard. If that’s not there yet, fix it first, because <strong>no tool is going to do that work for you</strong>.</p><p><em>This is the third article in a series about what AI adoption really looks like from the inside.</em></p><p><a href="https://medium.com/@stephaniegarmal/from-7-degrees-to-ai-a-nonlinear-career-path-f49f7f06d84b"><em>From 7 Degrees to AI: A Nonlinear Career Path</em></a><em> · </em><a href="https://medium.com/@stephaniegarmal/ai-in-the-real-world-why-legacy-codebases-are-the-true-test-of-ai-adoption-4942f9ec1f1d"><em>AI in the Real World</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=388b37ad88e2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI in the Real World: Why Legacy Codebases Are the True Test of AI Adoption]]></title>
            <link>https://medium.com/@stephaniegarmal/ai-in-the-real-world-why-legacy-codebases-are-the-true-test-of-ai-adoption-4942f9ec1f1d?source=rss-279b17d719c5------2</link>
            <guid isPermaLink="false">https://medium.com/p/4942f9ec1f1d</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[tech-leadership]]></category>
            <category><![CDATA[legacy-code]]></category>
            <dc:creator><![CDATA[Stephanie Garmal]]></dc:creator>
            <pubDate>Tue, 24 Feb 2026 14:53:26 GMT</pubDate>
            <atom:updated>2026-02-24T14:54:17.474Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Everyone’s talking about building with AI. Almost nobody’s talking about what happens when your codebase is older than the AI tools themselves.</em></p><h3>Starting from scratch is a different game</h3><p>When you build something new, even without AI, it’s just easier. You’re the one deciding how the architecture works, what libraries to use, and how everything connects. There’s no history to deal with. AI fits naturally into that because the rules are clear and consistent from day one, and honestly, that’s the version of AI adoption most people are writing about online.</p><p>But that’s not the reality for a lot of teams.</p><h3>The mess that comes with time</h3><p>When a product has been in production for years, the code comes with baggage. The most recent code might follow solid, modern patterns. But go deeper into the same project, and you’ll find older parts built with a completely different philosophy, different dependencies, sometimes a completely different way of thinking. Nobody went back to refactor those parts because they didn’t need it. They’ve been running fine in production, and touching them would be a risk without reward.</p><p>People who’ve been working in that code long enough know which parts to follow and which to leave alone. But when AI reads the same codebase, it treats all of it as equally valid.</p><h3>So what do you actually tell AI to follow?</h3><p>And that’s the question that catches most teams off guard. You can’t just hand AI the whole repo and say “<em>match the style</em>” because there isn’t one style. There’s a patchwork of decisions made by different people over different years. AI will absorb all of it, the modern approach and the deprecated one sitting three folders away, and it won’t know which one you’d want going forward.</p><p>You can ask AI to review the code and flag inconsistencies, and it’ll do a decent job at that. But deciding which pattern becomes the standard? That’s an architectural decision, not something you can automate.</p><h3>What this really comes down to</h3><p>The teams that actually succeed with AI in mature products won’t be the ones who picked the best tool. They’ll be the ones who did the less exciting work of figuring out their own standards first, deciding what AI should and shouldn’t learn from, and treating AI adoption as a team effort. Because it’s not about one person deciding what’s right. It’s about the whole team learning how to train AI with their feedback, correcting it when it picks up the wrong patterns, and building that discipline together.</p><p>That’s what AI adoption looks like when the codebase has history. Not a clean demo on a fresh repo, a real decision, on a real Monday morning, about which version of your own code gets to be the truth.</p><p><em>This is the second article in a series about </em><strong><em>what AI adoption really looks like from the inside</em></strong><em>. The first one, </em><a href="https://medium.com/@stephaniegarmal/from-7-degrees-to-ai-a-nonlinear-career-path-f49f7f06d84b"><em>From 7 Degrees to AI: A Nonlinear Career Path</em></a><em>, covers how my nonlinear background ended up being exactly what AI product leadership needs.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4942f9ec1f1d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From 7 Degrees to AI: A Nonlinear Career Path]]></title>
            <link>https://medium.com/@stephaniegarmal/from-7-degrees-to-ai-a-nonlinear-career-path-f49f7f06d84b?source=rss-279b17d719c5------2</link>
            <guid isPermaLink="false">https://medium.com/p/f49f7f06d84b</guid>
            <category><![CDATA[career-advice]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Stephanie Garmal]]></dc:creator>
            <pubDate>Wed, 18 Feb 2026 18:10:13 GMT</pubDate>
            <atom:updated>2026-02-19T04:25:23.918Z</atom:updated>
            <content:encoded><![CDATA[<p><em>How studying criminology, graphic design, and operations research prepared me for AI better than any bootcamp could.</em></p><p>When people ask how I ended up leading AI products, they expect a clean story. That’s not what happened.</p><p>I have seven university degrees, and none of them are in AI. Let me walk you through it.</p><h3>It started with two things at once</h3><p>I enrolled in Graphic Design and Computer Science simultaneously. Not because I was overachieving, but because they complement each other perfectly. Design taught me how people interact with interfaces. Computer Science taught me how to build what’s behind them. I’ve carried that dual lens ever since: what does the user need, and how do we make it work?</p><h3>I didn’t want to stay a developer</h3><p>I’ll be honest. Design and code were great, but I pursued Operations Research and Operations Management because I wanted to lead, not just build. I saw too many talented developers who stayed in execution for their entire career. Deep experts in one language, one framework, one corner of a system. That’s a valid path, but it wasn’t mine.</p><p>I wanted to be the person who sees the whole picture. Who defines <em>what</em> we build, not just <em>how</em>. Operations Research gave me the language of optimization, constraints, and decision-making under uncertainty. Over sixteen years later, that’s exactly what AI product leadership requires every single day.</p><h3>The cybersecurity detour that wasn’t a detour</h3><p>I wanted to study criminal technology, the kind you see in movies where someone traces a digital footprint and takes down a threat. That career didn’t exist in Guatemala, so I studied Criminalistics and Criminology first, then won a full scholarship from the Government of India to train in cybersecurity at C-DAC alongside professionals from 20 countries. I placed second overall.</p><p>That “detour” taught me something critical: every system has vulnerabilities. Today, when I evaluate AI products, I think about adversarial inputs, data poisoning, and misuse. Not from a textbook, but from training that forced me to think like an attacker.​​​​​​​​​​​​​​​​</p><h3>A generalist in a world that rewards specialists</h3><p>All of that led to a resume that looks scattered until you see the pattern: data science consultant for the EU during COVID, cloud engineer building a demographics product, senior backend developer on a health-tech platform, technical lead for a backend team, product lead for several projects across industries.</p><p>I’ve written Python, Java, R, and SQL. I’ve worked across Azure, GCP, and AWS. I’ve managed Kubernetes deployments and built data pipelines.</p><p><strong>I’m not the world’s best at any one of those things. And that’s the point.</strong></p><p>AI doesn’t need another deep specialist. It needs people who can talk to the C-suite about strategy and then walk into a standup and understand why the model isn’t performing. People who can bridge product stakeholders, design, and technical implementation. As one of my former VPs once wrote about me: <em>technical, detail oriented and personable, with a skillset not often found in the tech sector. The kind of person who connects the dots between vision and execution</em>. And this is exactly the kind of profile AI needs right now.</p><h3>What this means for AI</h3><p>The biggest barrier to AI adoption isn’t the technology. It’s the gap between the people building it and the people deciding to use it.</p><p>Business leaders are buying AI tools they don’t understand. Technical teams are shipping AI features that don’t solve real problems. And almost nobody is asking: what happens when this model is wrong? Who’s accountable?</p><p>That’s where a nonlinear background becomes an advantage. I can sit in a boardroom and speak strategy. I can sit with engineers and speak about architecture. I can think about risk because I studied how systems, both human and digital, break. And if your background looks anything like mine, that’s not a weakness.</p><h3>Your nonlinear background is your edge</h3><p>If you’re reading this and thinking your career path is too scattered for AI, it’s not. AI needs people who understand healthcare, logistics, finance, law, and education. The person who studied supply chain and then learned machine learning will build better AI for logistics than someone who only knows algorithms.</p><p>Your unusual path isn’t a detour. It’s your differentiator.</p><p>I’m proof.</p><p><em>Visit </em><a href="https://stephaniegarmal.ai/"><em>stephaniegarmal.ai</em></a><em> </em>| Connect on <a href="http://www.linkedin.com/in/stephaniegarmal">LinkedIn</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f49f7f06d84b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>