<?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 ChrisP on Medium]]></title>
        <description><![CDATA[Stories by ChrisP on Medium]]></description>
        <link>https://medium.com/@apps4cloud.eu?source=rss-613db88e05bf------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*_5a9gcWA5ZR3Q7z4aJNiFQ.png</url>
            <title>Stories by ChrisP on Medium</title>
            <link>https://medium.com/@apps4cloud.eu?source=rss-613db88e05bf------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 18:34:17 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@apps4cloud.eu/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[The Power Drill Doesn’t Make You a Craftsman]]></title>
            <link>https://medium.com/@apps4cloud.eu/the-power-drill-doesnt-make-you-a-craftsman-10cbc1412fd1?source=rss-613db88e05bf------2</link>
            <guid isPermaLink="false">https://medium.com/p/10cbc1412fd1</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[software-architecture]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[ChrisP]]></dc:creator>
            <pubDate>Fri, 15 May 2026 14:01:00 GMT</pubDate>
            <atom:updated>2026-05-15T14:01:00.719Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qjIlHWTZdbytLKTZOcEpNw.png" /><figcaption><strong>The Power Drill Doesn’t Make You a Craftsman</strong></figcaption></figure><p><strong>The Power Drill Doesn’t Make You a Craftsman</strong></p><p>I watched a video by Karell Ste-Marie recently, and one idea stayed with me: AI can help developers move faster, but it can also quietly stop junior and mid-level engineers from learning the most important skill: <strong>how to think through a problem</strong>.</p><p>That connected with something, I’ve been saying to my team for a while. Yesterday, during a 1-on-1, the metaphor finally clicked.</p><p>AI is a power drill.</p><p>Give a power drill to an experienced carpenter, and they’ll build something faster, cleaner, and with more precision. Give the same drill to someone who has never worked with wood, and they may just make mistakes faster. Crooked holes. Split timber. Go back. Fix things. And finally, an ‘ok’ furniture…</p><p>The tool didn’t change. The person using it did.</p><p>I’ve seen the same thing in software. A senior engineer can use AI to speed up complex work because they already know what to look for. They understand the domain, the security risks, the edge cases, and the parts that can break later. Someone without that experience might use the same AI tool, get a lot of code quickly, and still miss the important stuff.</p><p>Same drill. Very different result.</p><p>That’s the trap with AI. It feels productive. Code appears. Estimates are suggested. Architecture ideas are generated. Everything moves fast. But speed is not the same as understanding.</p><p>If you don’t question what AI gives you, you’re not really building software. You’re just connecting pieces and hoping they work together. And sure, by the end of the day they might fit, but the real question is: do they fit the problem you actually need to solve?</p><p>And this is why the learning years still matter. Maybe they matter even more now. AI can handle a lot of the mechanical work, but it can’t replace judgment. Judgment comes from experience. From debugging broken systems. From shipping things that didn’t go as planned. From learning why a pattern exists, not just copying it because it looks right.</p><blockquote>So yes, use AI. Use it a lot. It is a powerful tool.</blockquote><p>But don’t let it <strong>think</strong> for you.</p><p><em>Before you accept an answer, ask what it might be missing. Before you trust an estimate, think about the unknowns. Before you commit the code, understand the business flow, the risk, and the failure points.</em></p><p>The developers who will grow the most are not the ones who paste prompts the fastest. They are the ones who bring their own judgment to the table.</p><p>Become the craftsman!</p><p>The power drill will always be there!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=10cbc1412fd1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Is AI helping us ship faster, or are we trading speed for technical debt?]]></title>
            <link>https://medium.com/@apps4cloud.eu/is-ai-helping-us-ship-faster-or-are-we-trading-speed-for-technical-debt-2e422fe043cc?source=rss-613db88e05bf------2</link>
            <guid isPermaLink="false">https://medium.com/p/2e422fe043cc</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[software-engineering]]></category>
            <dc:creator><![CDATA[ChrisP]]></dc:creator>
            <pubDate>Tue, 12 May 2026 13:35:21 GMT</pubDate>
            <atom:updated>2026-05-12T13:35:21.837Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cXzPz08uoH9ADrNa-4Sc0g.png" /><figcaption>Is AI helping us ship faster, or are we trading speed for technical debt?</figcaption></figure><p>I am not against AI.</p><blockquote>Quite the opposite… Who does not use an electric screwdriver nowadays?</blockquote><p>I use AI, I learn from it, and I believe it can be an incredible force multiplier for software engineering teams.</p><p>What I mean is this: better tools are not the problem. The real question is whether we understand what kind of work the tool is helping us do, where it creates leverage, and where human judgment still matters.</p><p>But I also think we need to be brutally honest about something:</p><p>— AI does not automatically make software development better.</p><p>— It makes software development faster.</p><p>And faster is not always better.</p><p>Recently, I watched a <a href="https://www.youtube.com/watch?v=fvqD83ffMnw">video</a> by <a href="https://www.youtube.com/@theseriouscto">The Serious CTO</a> , Karell Ste-Marie, about technical debt and how developers often introduce it while trying to ship features quickly. The pattern is familiar to most engineering teams: we need to release an MVP, validate an idea, satisfy a customer, hit a deadline, or unblock the business. So we make compromises…</p><p>— A shortcut here.</p><p>— A missing abstraction there.</p><p>— A skipped test.</p><p>— A quick conditional.</p><p>— A TODO that everyone silently knows will probably never be revisited.</p><p>Each decision feels small and reasonable in isolation. But over time, these decisions compound. The codebase becomes harder to understand, harder to change, harder to test, and harder to trust. Eventually, the team realizes that every new feature takes longer than expected. Delivery slows down. Estimates become unreliable. Engineers become more cautious. The system starts pushing back.</p><p>Technical debt does not usually arrive as one big catastrophic decision.</p><p>It arrives as a thousand small tactical decisions.</p><h3>AI as a Tactical Tornado</h3><blockquote>This is where AI changes the equation.</blockquote><p>A friend and colleague, <a href="https://medium.com/u/11268a56bad5">Yannis Rizos</a>, described this very well in a recent <a href="https://medium.com/@yrizos/the-fastest-engineer-in-the-room-44168a7570b8">post</a>, connecting it to what John Ousterhout describes as the problem of human tactical tornadoes: developers who move fast, produce a lot, but often leave complexity behind for others to clean up.</p><p>But now, the AI era has brought another tornado to the surface.</p><p>AI itself.</p><p>AI can generate code very quickly. It can scaffold a feature, write a test, suggest a refactor, explain an error, produce boilerplate, and help us move through problems that previously required much more manual effort.</p><p>That is powerful.</p><p>But it also introduces a new risk.</p><p>AI can become a tactical tornado.</p><p>It can move fast, produce a lot, and appear productive, while lacking deep understanding of the system, the business context, the architectural constraints, the historical decisions, and the long-term maintenance cost.</p><p>The danger is not that AI writes bad code by default.</p><p>The danger is that AI can produce code that looks good enough to merge, as Karell mentions on his YouTube <a href="https://youtu.be/fvqD83ffMnw?si=0dQUJ77v9IzJ_pp9&amp;t=215">video</a>.</p><p>And “good enough to merge” is not the same as “good enough to maintain.”</p><p>This is where John Ousterhout’s <em>A Philosophy of Software Design</em> becomes highly relevant. One of the strongest ideas in the book is the distinction between tactical and strategic programming.</p><blockquote>Tactical programming focuses on getting something working as quickly as possible.</blockquote><blockquote>Strategic programming accepts that good design requires thought, patience, and investment. It recognizes that the fastest way to build software in the long term is not always the fastest path in the next hour.</blockquote><p><em>AI tempts us toward tactical programming at scale.</em></p><p>And that should make us uncomfortable.</p><h3><strong>We may be buying time today and paying interest tomorrow</strong></h3><p>The common argument in favor of AI-generated code is simple:</p><blockquote>“Even if AI creates some mess, AI will help us fix it later.”</blockquote><p>Maybe.</p><p>But I think this argument is weaker than it sounds.</p><p>To fix something later, AI still needs <em>context</em>. It needs to understand what the system <em>is supposed to do</em>, why certain <em>decisions</em> were made, what <em>trade-offs</em> were accepted, what <em>constraints</em> exist, and what should <em>not be changed</em>.</p><p>If that context was never captured, reviewed, or understood by the team, then the cleanup is not free.</p><blockquote>It becomes expensive.</blockquote><p>Not only in engineering time, but also in <em>tokens</em>, <em>prompts</em>, <em>reviews</em>, <em>debugging sessions, and rework</em>.</p><p>We may be buying time today and paying it back later with interest.</p><p>And in the AI era, that interest may be paid in burned tokens, lost context, production incidents, and late-night debugging.</p><p>This is not a small problem. In the same video, Karell on his “The Serious CTO” youtube channel, referenced the staggering financial impact of technical debt. Industry reports have estimated the cost of poor software quality in the U.S. at trillions of dollars, with accumulated technical debt alone measured at more than a trillion dollars.</p><p><strong>Now add AI to that equation.</strong></p><p>If we use AI to generate more code faster, but without enough review, ownership, or architectural discipline, we may not only increase the traditional cost of technical debt. We may also introduce <strong>a new operational cost</strong>: the cost of repeatedly spending tokens to rediscover context, explain unclear code, generate fixes, validate assumptions, and repair decisions that should have been reviewed properly in the first place.</p><h3>And now the 2 AM problem…</h3><p>Imagine shipping a feature quickly with the help of AI. The code works, the tests pass (assuming again that tests exist), the pull request looks acceptable, and the feature goes live. Then, weeks later, production breaks at 2 AM. You open the code and realize that nobody fully understands it. The intent was not documented, the trade-offs are unclear, the edge cases are hidden, and now you are depending on AI again to explain the same code it helped produce. Only this time, the system is down, users are affected, the business is waiting, and in the worst possible moment, <em>you realize you have run out of tokens</em>. Dramatic? Maybe. But the point is real: if no human deeply reviewed the code, understood the design, captured the reasoning, and took ownership, the team did not really move faster. It deferred understanding. And deferred understanding is one of the most dangerous forms of technical debt.</p><h3>AI should be used strategically, not tactically</h3><blockquote>The answer is not to avoid AI</blockquote><p>That would be unrealistic and, frankly, a missed opportunity. The answer is to use AI strategically: to explore options, challenge assumptions, generate alternatives, explain unfamiliar code, write first drafts, speed up repetitive work, and improve tests, documentation, and refactoring plans. But we should not outsource engineering judgment, ownership, design thinking, or the responsibility of understanding what is being shipped. AI should help engineers think better, not help teams avoid thinking. That distinction matters.</p><h3>Guardrails Matter More Than Ever</h3><p>If AI increases <em>development speed</em>, then our engineering guardrails need to become <strong>stronger</strong>, not weaker.</p><p>Otherwise, <strong>we are simply accelerating entropy</strong>.</p><h3>Is Specification-Driven Development enough?</h3><blockquote>Specification-Driven Development can help.</blockquote><p>Clear specs give AI better context. They reduce ambiguity. They make expectations explicit. They help teams align before implementation starts.</p><p>But specs alone are not enough.</p><p>A specification can describe what needs to be built, but it does not automatically guarantee good design.</p><p>It does not guarantee maintainability.</p><p>It does not guarantee simplicity.</p><p>It does not guarantee that the implementation fits the architecture.</p><p>It does not guarantee that the team understands the code.</p><p>Specification-Driven Development is useful, but it should be part of a broader engineering discipline.</p><p>Specs help AI produce better output.</p><p><strong>Design thinking helps engineers decide whether that output should exist in the first place.</strong></p><h3>AI is an accelerator. Direction still matters.</h3><blockquote>AI is not the problem.</blockquote><p><em>Undisciplined usage is the problem.</em></p><p>AI can absolutely help us build better software. But only if we use it with engineering discipline.</p><p>The goal should not be to generate the maximum amount of code in the minimum amount of time.</p><p>The goal should be to create useful, understandable, maintainable systems faster than before.</p><p>That requires a <strong>strategic mindset</strong>.</p><p>It requires <strong>guardrails</strong>.</p><p>It requires <strong>review</strong>.</p><blockquote>It requires ownership.</blockquote><p>It requires engineers who are willing to slow down at the right moments so the team can move faster later.</p><p>Because the real question is not whether AI can help us ship faster.</p><p><strong>It can.</strong></p><p>The real question is whether we are using AI to reduce complexity or to generate complexity faster than we can understand it.</p><p>And if we are not careful, tomorrow’s technical debt will not come only from rushed human decisions.</p><p>It will come from rushed AI-assisted decisions that nobody truly reviewed, nobody fully understood, and nobody documented.</p><p>AI may help us buy time.</p><p>But without discipline, <em>that time may become very expensive to repay</em>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2e422fe043cc" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>