<?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 Kristian Trenskow on Medium]]></title>
        <description><![CDATA[Stories by Kristian Trenskow on Medium]]></description>
        <link>https://medium.com/@trenskow?source=rss-70e5a6904530------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*77myGzgpSfyfDBUB.jpg</url>
            <title>Stories by Kristian Trenskow on Medium</title>
            <link>https://medium.com/@trenskow?source=rss-70e5a6904530------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 12:42:14 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@trenskow/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[Why coding still matters in an AI world]]></title>
            <link>https://medium.com/@trenskow/lets-talk-about-me-being-unemployed-in-an-ai-world-973635ef03e8?source=rss-70e5a6904530------2</link>
            <guid isPermaLink="false">https://medium.com/p/973635ef03e8</guid>
            <category><![CDATA[coding]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[unemployment]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Kristian Trenskow]]></dc:creator>
            <pubDate>Mon, 04 May 2026 16:38:01 GMT</pubDate>
            <atom:updated>2026-05-06T09:19:10.562Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6EpaeYqgtQnJwb8iAPUGIA.png" /></figure><p>Last week I wrote a post on LinkedIn where I complained about being without work for a year.</p><p>For those who don’t know me: I’m a software engineer with 15 years of experience with iOS and Android. I have architectural experience and have also been a tech lead on multiple projects.</p><p>I got a lot of traction and feedback on my post about my current employment status (or lack thereof) — and it felt like some much needed attention in the absence of substance in my current daily life.</p><p>That being said, my long unemployment is also a bit of a self-inflicted wound. I’ve been a little too picky about my options and that is primarily what has let to my situation — so the post was mostly an attempt to gain some reach on some sympathy I really didn’t deserve.</p><p>But the whole sitation brought up some interesting thoughts, that I thought I’d like to share. A lot of people came back to me and mentioned that the fact I don’t have AI on my resume might be a reason why employers find me unattractive.</p><h3>So let’s talk about AI…</h3><p>First of all, everybody uses AI. AI is a great tool that enhances any developer — and all developers are using it. To which extend each individual developer is using it might differ, but I think I can guarantee that 99.9% of developers has embraced AI.</p><p>So to me being rejected because of the lack of AI experience on my resume sound to me the same as being rejected for not having explicitly written that I have experience with Xcode — even though it says iOS developer. At that standard I wouldn’t be surprised if I was rejected from a job opening because my resume didn’t explicitly say that I have experience using a Mac.</p><p>That of course is a bit black and white — but you get the gist. To me the tools has never been interesting. To me the products, the results, is what is interesting. What I am able to create is what is interesting — the means to an end is an interesting discussion but it’s ultimately not what matters.</p><p>That being said I of course understand the need for employers to know to which extend you’re using AI. There’s a long way from just casually using ChatGPT or GitHub Copilot to using Claude Code for your every task. But every developer lives in that span somewhere — but just writing “AI” on my resume does not change the perception of where I am on that slope.</p><p>So I thought I’ll write up this to give you some idea of where I stand on AI.</p><p>As I mentioned above I think AI are great tools for enhancing developer skills — and by AI I of course mean LLMs (even though AI in general is a much broader term).</p><p>To be honest I’m at the beginning of the slope. I’m not a Claude Code kind of guy. To me reasoning about problems and finding elegant solutions is what makes coding great. Delegating that to an AI is makes it hard to maintain those skills.</p><h3>Why coding matters</h3><p>But that does not mean that I don’t think tools like Claude Code has its merits. I think they are actually great tools for a subset of tasks. But there are places where it in my opinion just doesn’t fit. It’s in the code that matters. It is in the user experience that matters. Sure, go let Claude Code make that settings screen in your app, but the screen where you users actually spend their time must be written by a human. Why? Because it’s the details that matters. It’s the way the user’s touch just snaps the interface. The way it just feels nice and pleasant to interact with.</p><p>To me that is because there are some things that cannot be expressed in human language. That’s why you can try to explain something you are trying to achieve to Claude Code, but no matter how hard you try it will come back with “Oh, now I understand…” — only to get it wrong for the 20th time.</p><p>That’s because code is not just language. Code is a complex mathematical construct that allows you to describe the mental modal in your head. It’s like asking an LLM to describe how elections move around a nucleus. It’s like asking how electricity flows in a circuit. It’s like asking it to describe Einstein’s equations on gravity. There is no human language that is able to convey that entirely.</p><p>Yes, an LLM will give you an answer — it might even give you some nice illustrations to help you understand how gravity works — but you cannot really fathom how gravity works until you have gained the tools that can help you build the mental model you need to fully understand how it works. And those tools are not linguistic. In the case of Einsteins equations on gravity they are mathematical tools that do not have a one-to-one mapping to language. The LLM might even understand it on an abstract level — but it also only has human language to convey it across the screen.</p><p>And the same is true the other way around. There are just some things you cannot describe in language and that you therefore cannot make an LLM understand. Even if the LLM and you might have a shared mental model (even though LLMs are not mental) — the natural language barrier is still between you. The natural language becomes the bottleneck.</p><p>So maintaining your mental model about computer science problems becomes vital if you want to create actual good, safe and reliable user experiences. There are just some things that can ONLY be expressed in code.</p><p>And that’s where I find Claude Code a little dangerous also. It’s so addictive. Once you’ve learned that you can make Claude Code do simple tasks for you it becomes your goto tool for everything. But maintaining your mental model is only achieved by actually writing code daily and having the model present in your head at all time. Delegating work to Clause Code makes you lose that ability — and in my experience that deterioration starts pretty quickly.</p><h3>What is going to happen?</h3><p>That’s why it’s of utmost importance to me to keep coding. Sure, I can make Claude Code make that settings screen that only about 5% of users venture into. But the core elements of the products that users cares about needs have been written by humans who actually understood how to describe and solve the problem in code — not language. Humans that can feel how it is supposed to work and who has the tools to describe that feeling in code.</p><p>More and more experienced developers fall into the agentic coding trap — and they will in time forget how to be good software engineers — but they will become good prompters. Don’t get me wrong… there is nothing wrong with being good at prompting — but I’m convinced that once the hype has settled the actual skill of being able to reason about code is gonna be much more valuable than prompting skills — because let’s be honest… prompting is not hard — coding is hard.</p><p>And that’s why I want to stay in that lane. I’m convinced that being a good software developer with exceptional coding skills will be a most valuable commodity in the future. That might make me unemployed for the time being when everybody is rushing to agentic coding — but maybe then unemployment is a badge of honor right now — staying away from it all might be the best thing I’ll ever do career-wise. I just hope the business hasn’t forgot about me when my to be super powers are needed.</p><h3>To wrap up…</h3><p>Yes, I use AI. I use GitHub Copilot a lot. To be honest I cannot live without Copilot — even though I turn it off sometimes to force myself to write actual algorithms — but I mostly sit patiently with my fingers eager on the keyboard in anticipation of Copilot’s next suggestion. I use ChatGPT (and Claude) for asking questions that I don’t want to spend hours googling. I use my skills to filter whatever response I get — and the LLM and I actually make a pretty constructive partnership.</p><p>But I do still write most of the code myself. I still feel like I need to be in control. I need to know the exact reasoning behind every single line of code I write. Every line of code needs a justification to be there — and letting an LLM spit out 10.000 lines of code in a day is guaranteed to be at least 9.500 lines of code that has no justification to be there.</p><p>With all that said it might also look different for me in the future. I was slow to adopt ARC when it was implemented in Objective-C because I didn’t trust it. I was manually managing memory way longer than anyone else. Today I cannot live without ARC. I didn’t trust Swift when it came. I wrote Objective-C way longer than anyone else. Today I need a very good reason to not write my code in Swift (when targeting Apple platforms).</p><p>So who knows… maybe in a year or two I will also be agentic coding — but to be honest… I don’t think so. I might be using it more than I do now but I do not think I’ll ever write entire applications in Claude Code.</p><p>The reason?</p><p>I’ll never get over the language barrier.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=973635ef03e8" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>