<?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 Chris Lester on Medium]]></title>
        <description><![CDATA[Stories by Chris Lester on Medium]]></description>
        <link>https://medium.com/@chrisglester?source=rss-6dbacd26aff2------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*KX0rZlPOxx7kPVSnlhM32g.jpeg</url>
            <title>Stories by Chris Lester on Medium</title>
            <link>https://medium.com/@chrisglester?source=rss-6dbacd26aff2------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 02:46:00 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@chrisglester/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 First Time AI Followed Me to Work]]></title>
            <link>https://medium.com/@chrisglester/the-first-time-ai-followed-me-to-work-23efb87af321?source=rss-6dbacd26aff2------2</link>
            <guid isPermaLink="false">https://medium.com/p/23efb87af321</guid>
            <category><![CDATA[beginnerjourney]]></category>
            <category><![CDATA[helpdesk]]></category>
            <category><![CDATA[co-pilot]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[productivity]]></category>
            <dc:creator><![CDATA[Chris Lester]]></dc:creator>
            <pubDate>Sun, 15 Nov 2026 16:22:12 GMT</pubDate>
            <atom:updated>2026-11-15T16:22:12.879Z</atom:updated>
            <content:encoded><![CDATA[<p>Where my personal AI experiments finally collided with a real workplace problem — and what happened next.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*hDf-wx3_aYSvD5yu" /><figcaption>Photo by <a href="https://unsplash.com/@glenncarstenspeters?utm_source=medium&amp;utm_medium=referral">Glenn Carstens-Peters</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Up until now, every AI project I’ve written about has lived safely in my personal world — late‑night tinkering, messy prototypes, and the occasional geometry riot. But eventually curiosity leaks into the rest of your life, and for me, that moment arrived the day a Helpdesk ticket pushed me over the edge.</p><p>It had no information.<br>It was already old.<br>And it was blocking a finance user’s workflow.</p><p>I’ve been at this company for eleven years. I started as the only Helpdesk tech — answering every call, manually creating every ticket, and sitting in weekly meetings where I walked leadership through the entire queue one ticket at a time. So when a ticket arrives with zero troubleshooting, zero notes, and zero effort… let’s just say I have opinions.</p><p>That frustration turned into a question:<br><strong>Could AI help the Helpdesk do the job the Helpdesk is supposed to do?</strong></p><h3>The First Build: A Helpdesk Assistant</h3><p>I’d heard about Claude Skills and figured I could build something similar. Copilot’s enterprise version has its own skills‑style feature, so I started experimenting.</p><p>The idea was simple:<br>Give the Helpdesk a small assistant they could paste ticket contents into — or type their own explanation — and get guided questions, basic troubleshooting, and enough context to avoid sending empty tickets downstream.</p><p>A little tweaking, and suddenly I had something that actually looked good. Clean. Straightforward. Useful.</p><p>Later, I thought of another feature and added it.<br>Then I tried recreating the same assistant in Claude on my personal account, mostly out of curiosity. Selling it wasn’t realistic — those YouTubers host their skills behind paywalls on their own sites — and I’m not spinning up infrastructure for a simple helper tool.</p><p>So I focused on the version that mattered: the one for work.</p><h3>The First Real Test</h3><p>My manager gave me the green light to show the assistant to the Helpdesk managers.</p><p>Of course, right in the middle of this, Copilot lost my agent.<br>Gone.<br>Poof.</p><p>Life lesson: always save a backup.<br>I didn’t.</p><p>So I rebuilt it from scratch — faster this time, because I knew exactly how I wanted it to work — and saved the markdown file locally.</p><p>I was able to present this to one of the Helpdesk managers at the tail end of another meeting. He was receptive but skeptical. Our ticketing system is being upgraded soon, and the new version supposedly has AI built in. He wasn’t convinced it would work, and he thought everyone needed a license to use an agent.</p><p>We went back and forth — I don’t have a license, my teammate doesn’t have a license — and eventually we agreed to share it with one Helpdesk tech and pilot it for a short period.</p><p>That was the moment AI officially crossed the boundary between my personal projects and my day job.</p><h3>Meanwhile, the Trip App Keeps Evolving</h3><p>Even while the Helpdesk assistant was being tested, I kept working on the trip board app.</p><p>I expanded the AI planning tab to generate full‑day itineraries.<br>I asked Claude to finalize build info before moving to Claude Code.<br>I had Claude reviewing the app for bugs, security issues, and incomplete logic.<br>I explored cost models and publishing options.<br>Eventually Claude packaged everything up — and reminded me that Claude Code requires a subscription, along with all the other infrastructure pieces I’d need.</p><p>It’s been a slow, steady grind. But the app is becoming real.</p><h3>The Pilot Results</h3><p>When I checked in with the Helpdesk tech who tested the assistant, her feedback was exactly what I hoped for:</p><p><strong>Overall helpful.</strong><br>Some situations didn’t apply — which is expected — but the core idea worked.</p><p>I sent her notes and findings along. Now it’s a matter of seeing where this leads.</p><h3>What This Moment Means</h3><p>This wasn’t a dramatic workplace transformation. It wasn’t a sweeping automation project. It was a small, focused attempt to solve a real problem — and it worked well enough to matter.</p><p>More importantly, it marked a shift in my journey.</p><p>For months, AI has been something I explored alone — building apps, experimenting with game logic, pushing tools until they broke. But this was the first time those skills crossed into my professional world.</p><p>It wasn’t planned.<br>It wasn’t polished.<br>It wasn’t perfect.</p><p>But it was real.</p><p>And it showed me something I didn’t expect:<br><strong>the line between my personal AI projects and my day‑to‑day job is thinner than I thought.</strong></p><p>This post marked the moment my tinkering crossed into my day job.<br>I’m not sure yet what the next post will be about. While I haven’t purchased the subscription to move forward with my travel app, and work and life have been busy, I still have smaller projects I can cover. Or perhaps I’ll take a step back and talk about the potential I see for AI in business.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=23efb87af321" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Mirror of Criticism: Breaking the AI “Yes-Man” Loop]]></title>
            <link>https://medium.com/@chrisglester/the-mirror-of-criticism-breaking-the-ai-yes-man-loop-e88505d3b292?source=rss-6dbacd26aff2------2</link>
            <guid isPermaLink="false">https://medium.com/p/e88505d3b292</guid>
            <category><![CDATA[tinkerjourney]]></category>
            <category><![CDATA[promtengineering]]></category>
            <category><![CDATA[ai-coding]]></category>
            <dc:creator><![CDATA[Chris Lester]]></dc:creator>
            <pubDate>Mon, 09 Nov 2026 19:28:04 GMT</pubDate>
            <atom:updated>2026-11-09T19:28:04.600Z</atom:updated>
            <content:encoded><![CDATA[<p>Transitioning from passive text generation to active architectural design requires recognizing when an AI model is acting as a “yes-man”. True development momentum begins when you deliberately prompt for harsh, objective criticism to bridge the gap between a visual mockup and a functional application.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*O-TVGB4viBjSGo3S" /><figcaption>Photo by <a href="https://unsplash.com/@silverkblack?utm_source=medium&amp;utm_medium=referral">Vitaly Gariev</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Every tinkerer knows the quiet comfort of working in isolation. You can spend an entire weekend tweaking the visual aesthetics of a personal project — like trying to make a digital travel app look like a warm, retro suitcase outline or a set of luggage tags — without ever confronting whether the thing actually works. For a long time, my AI coding companions were happy to indulge me. They nodded along, built whatever interface elements I threw at them, and confidently told me “sure, no problem” every single time I pushed for a new feature.</p><p>But this past week, I realized something dangerous about building with artificial intelligence: by default, AI is a chronic “yes-man”. If you don’t intentionally break that loop, you can spend days building a beautiful, polished prototype that is completely hollow on the inside.</p><p>Here is what happened when I stopped asking my AI assistant to be helpful and told it to start being brutal.</p><h3>The Stalemate of Two Platforms</h3><p>For days, I found myself trapped in an agonizing loop of indecision with my vacation trip board organizer — a project I am still trying to find the perfect name for. I was bouncing back and forth between two entirely different development tracks, trying to bring both versions to the same state:</p><ul><li><strong>The No-Code Track (Base44):</strong> This version looked great out of the box, but I hit a hard wall when I discovered that adding a simple Google Search API to improve search speed and accuracy required backend functions locked behind a steep monthly builder subscription.</li><li><strong>The Native Code Track (Claude):</strong> This sandbox handled my complex data instructions much better and didn’t charge me for basic API integrations, but the visual interface looked incredibly dated. Even after I forced it to completely rebuild the architecture to work smoothly on mobile phones, I wasn’t sold on the look.</li></ul><p>I was jumping back and forth across browser tabs, completely losing track of what constraints I had given to which model. Frustrated by the visual and functional stalemate, I decided to try an experiment. Instead of asking Claude to build another feature, I asked it to give me a completely unvarnished, harsh critique of the app we had built together so far.</p><p>And boy, it did not hold back.</p><h3>Surviving the Digital Roast</h3><p>When you prompt an AI to criticize your work correctly, it stops acting like a polite assistant and starts acting like a strict, technical auditor. Claude systematically picked apart the prototype.</p><p>Because it was a surface-level mockup, much of the underlying data logic didn’t actually work yet — which the AI was quick to point out (duh). But the real eye-opener came from the user-experience criticisms. The AI pushed a series of structural recommendations that completely clashed with my personal vision for how the app should feel.</p><p>It was a jarring but incredibly illuminating moment. I realized that if you leave an AI to its own devices, it will naturally drift toward the path of least resistance or generic, cookie-cutter design patterns. It will happily agree with your bad design choices just to give you a functional script.</p><p>Receiving that harsh feedback forced me to step up as the definitive product owner. It taught me a vital lesson about the human-AI relationship: you have to learn to accept valid technical critiques while actively fighting to defend your original creative intent.</p><h3>Pivoting the Strategy</h3><p>Learning to negotiate with the AI’s criticism also changed how I viewed my other major hobby project: the real-time strategy game inspired by <em>Battle for Middle-Earth II</em>.</p><p>After watching a legion of invisible circle-people overrun my fortresses in previous builds, I spent some time thinking about how to get the AI to generate meaningful graphic improvements. But looking at the project through a more critical lens, I had to face a reality check: finding a way to get Claude to generate the right graphics was going to be a journey I’m not sure I was prepared to struggle through.</p><p>Instead of abandoning the massive, detailed “world bible” that Copilot and I had engineered, I decided to pivot the entire delivery format. We began discussing how to take that rich lore, the factions, and the combat mechanics and repurpose them entirely. We are stepping away from a 2D digital simulation and transforming the project into a fully realized tabletop role-playing game (RPG) campaign book and rulebook.</p><h3>What I Learned on the Ground</h3><ul><li><strong>Beware the Complacency Loop:</strong> An AI will confidently build exactly what you ask for, even if it’s a terrible user experience. You have to explicitly ask for harsh feedback to break the “yes-man” cycle.</li><li><strong>The Execution Gap is Massive:</strong> It is incredibly easy to feel like a wizard when an AI generates a flashy daily itinerary builder featuring three-meal plans and localized activity tracking. But there is a massive difference between a visual mockup and a secure application ready for an app store.</li><li><strong>Frame the Constraints, Defend the Vision:</strong> If you don’t maintain a firm grip on the steering wheel, the machine will subtly redesign your project into something you didn’t want.</li></ul><p><strong>What’s Next?</strong></p><p>The sandbox is getting bigger, the critiques are getting tougher, and the projects are starting to feel real. I spent the weekend diving back into the travel app, pushing past the initial feedback to build out an advanced AI planning module that generates comprehensive multi-location itineraries, full three-meal daily structures, and localized activity tracking. Whether these personal projects end up as a published app store utility or a printed RPG rulebook, the momentum hasn’t slowed down.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e88505d3b292" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Game Dev: Building an RTS Game Prototype from Scratch]]></title>
            <link>https://medium.com/@chrisglester/ai-game-dev-building-an-rts-game-prototype-from-scratch-948f4e696804?source=rss-6dbacd26aff2------2</link>
            <guid isPermaLink="false">https://medium.com/p/948f4e696804</guid>
            <category><![CDATA[game-development]]></category>
            <category><![CDATA[online-rts-game]]></category>
            <category><![CDATA[ai-development]]></category>
            <category><![CDATA[claude-ai]]></category>
            <category><![CDATA[tinkerjourney]]></category>
            <dc:creator><![CDATA[Chris Lester]]></dc:creator>
            <pubDate>Thu, 15 Oct 2026 19:02:20 GMT</pubDate>
            <atom:updated>2026-10-15T19:02:20.095Z</atom:updated>
            <content:encoded><![CDATA[<h3>From Circles to Elves: The Real-Time Strategy Experiment</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*QvX_31_xS8tR-59W" /><figcaption>Photo by <a href="https://unsplash.com/@emily_wade?utm_source=medium&amp;utm_medium=referral">Emily Wade</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>My broader series began with a simple idea: <em>Beginning My Journey in AI.</em> Back then, I wrote, “I’m not an expert in artificial intelligence.” I was just curious. And curiosity has a habit of dragging you into increasingly strange and ambitious experiments.</p><p>My first major foray into building with AI ended with a vacation trip board that ran out of computing power and a database search that moved like molasses. But if there’s one thing you should know about being a tinkerer, it’s that a failed project isn’t a dead end — it’s a pivot point.</p><p>Wrestling with the strict limitations of no‑code app builders taught me something important: I needed a tool capable of handling complex logic without hiding the underlying machinery. That realization sent me down a YouTube rabbit hole, sparked a fresh curiosity about Claude’s coding environment, and led to a wild question:</p><p><strong>Could I actually build a playable video game using artificial intelligence?</strong></p><p>Here’s what happens when a non‑developer tries to engineer a real‑time strategy (RTS) prototype from scratch.</p><h3>Blueprinting the Battlefield with Copilot</h3><p>Before writing a single line of code, you need a design.</p><p>I didn’t want to make just any game — I wanted to recreate the tactical depth of <em>Battle for Middle‑Earth II</em>, one of the greatest RTS games ever made. Its squad‑based combat made battles feel massive and dynamic, rather than just clicking individual units one by one.</p><p>To bring this vision to life, I teamed up with Copilot to build a comprehensive <strong>world bible</strong>. Together, we mapped out:</p><ul><li><strong>Lore &amp; Factions:</strong> Distinct backstories and thematic identities.</li><li><strong>Gameplay Mechanics:</strong> Economy rules, base building, squad behavior, combat flow.</li><li><strong>Unit Specifications:</strong> Movement speeds, attack types, advantages, counters.</li></ul><p>With a massive text file packed with precise world‑building and mechanics, I felt ready. I downloaded the entire document, opened Claude, pasted the text, and issued a bold command:</p><p><strong>“Make an RTS based on this world bible.”</strong></p><p>Claude responded instantly and confidently: <em>“Sure, no problem.”</em></p><h3>The Red Square Rebellion</h3><p>When the code finished generating, I opened the prototype with pure anticipation. I expected the early stages of an epic fantasy battlefield.</p><p>Instead, I saw circles.</p><p>Circles everywhere.</p><p>Against a plain digital canvas, a few lonely squares sat motionless. Then, a massive swarm of little circle‑people materialized, marching in perfect, terrifying formation across the screen. Within minutes, this abstract geometry army overran and destroyed my fortresses — fortresses I couldn’t even see because Claude hadn’t rendered them yet.</p><p>It wasn’t <em>Lord of the Rings</em>.</p><p>It was a high‑speed geometry riot.</p><h3>The “Precision Gap” and What It Teaches You</h3><p>As I watched my invisible strongholds fall to a legion of circles, a crucial lesson about AI development hit home:</p><p><strong>AI will not always fill in the blanks.</strong></p><p>When you tell a generative model to “build an RTS,” it doesn’t intuitively know that you want:</p><ul><li>custom elven sprite animations</li><li>visible castle walls</li><li>a balanced resource economy</li><li>squad formations</li><li>terrain‑based movement rules</li></ul><p>Some models ask follow‑up questions. Many don’t. They choose the path of least resistance to produce a technically functional script.</p><p>If you aren’t exceptionally precise with your visual constraints, asset placeholders, and rendering rules, you get circles instead of elves.</p><p>To fix the geometry riot, I had to change how I talked to the technology.</p><p>I went back to Copilot — not to write lore, but to draft <strong>hyper‑specific technical parameters</strong>. We broke the game down into granular</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=948f4e696804" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Beginning My Journey in AI: Falling Into the First Rabbit Hole]]></title>
            <link>https://medium.com/@chrisglester/beginning-my-journey-in-ai-falling-into-the-first-rabbit-hole-82843836054a?source=rss-6dbacd26aff2------2</link>
            <guid isPermaLink="false">https://medium.com/p/82843836054a</guid>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[beginner]]></category>
            <category><![CDATA[microsoft-copilot]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[no-code]]></category>
            <dc:creator><![CDATA[Chris Lester]]></dc:creator>
            <pubDate>Wed, 16 Sep 2026 19:36:02 GMT</pubDate>
            <atom:updated>2026-09-16T19:36:02.295Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xc4ShRVJat909GuN" /><figcaption>Photo by <a href="https://unsplash.com/@guillaumeissaly29?utm_source=medium&amp;utm_medium=referral">Guillaume Issaly</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>“Beginning My Journey in AI” was the title of the first note I ever wrote about artificial intelligence. At the time, I admitted something simple: <em>“I’m not an expert in artificial intelligence.”</em> I was just curious — and curiosity has a way of pulling you into unexpected rabbit holes.</p><p>This post is the story of the first one.</p><p>While I had already experimented casually with AI tools, my journey truly shifted when I stopped just chatting with AI and tried to <strong>build something real</strong>. As a lifelong tinkerer, I wanted to see if I could construct a functional, customized application from scratch without going back to school for a computer science degree.</p><p>If you’re standing on the sidelines wondering how a non‑developer actually navigates these tools, here’s exactly what happened when I tried to build my first app.</p><h3>The Vision vs. The “No‑Code” Wall</h3><p>I started where most beginners do: a low‑barrier‑to‑entry platform that promised to turn simple text prompts into working software.</p><p>My goal was straightforward — a <strong>vacation trip board organizer</strong> to plan and visualize travels.</p><p>At first, it felt like magic. I typed in what I wanted and a clean, basic interface appeared. It felt like wizardry.</p><p>But the honeymoon phase ended fast. As soon as I began testing the app, I slammed into invisible walls:</p><ul><li><strong>The Sluggish Search:</strong> The database search was painfully slow, making it useless for filtering destinations or itineraries.</li><li><strong>The Credit Crunch:</strong> The AI‑powered features ran on background credits… which evaporated almost instantly.</li><li><strong>The “Free” Tier Wall:</strong> Before I could finish a prototype, the platform cut me off entirely.</li></ul><p>I wasn’t ready to abandon the project, but the tool had run out of resources long before I ran out of ideas.</p><h3>The “AI‑to‑AI” Troubleshooting Loop</h3><p>This is where the rabbit hole truly opened.</p><p>Determined to keep going, I bypassed the no‑code interface and started researching workarounds. YouTube, Google, forums — anything that could help me understand why the app was struggling.</p><p>That led me to a bizarre multi‑AI workflow that felt like orchestrating a digital team of rivals:</p><ol><li>Copy the sluggish search code or system logs from the app builder.</li><li>Paste them into Copilot and ask: <em>“Why is this tool processing searches so slowly, and how can we optimize the query?”</em></li><li>Take Copilot’s fix and drop it into another AI model: <em>“Can you rewrite this to use fewer computing credits?”</em></li></ol><p>Sometimes it felt brilliant — one AI acting as translator, another as debugger. Other times it was a comedy of errors. One AI would confidently propose a fix that relied on a feature the other AI didn’t support, creating a loop of digital finger‑pointing.</p><p>But even when it failed, I was hooked. I wasn’t just consuming technology anymore; I was negotiating with it.</p><h3>What I Learned on the Ground</h3><p>If you’re thinking about using AI to build your first app, skip the marketing hype and keep these practical realities in mind.</p><h3>1. No‑Code Doesn’t Mean No‑Logic</h3><p>AI can write syntax, but <em>you</em> still have to design the logic. If you don’t know how your data should flow, the AI will happily build a beautiful interface that leads nowhere.</p><h3>2. The Resource Trap Is Real</h3><p>“Free” tiers and credit systems are great for mockups, but they collapse the moment you start testing real logic or search capabilities.</p><h3>3. The Multi‑Tool Approach Wins</h3><p>Don’t rely on a single AI. Different models catch different mistakes. When you hit a wall, ask another AI to look at the problem from a fresh angle.</p><h3>What’s Next?</h3><p>My first vacation trip board was a messy, mismatched puzzle of tools — but it proved something important: <strong>the barrier to entry has officially collapsed</strong>. You don’t need to be an expert to start building; you just need patience and curiosity.</p><p>After wrestling with the limitations of basic app builders, I realized I needed a tool that could handle more complex, end‑to‑end coding without forcing me into a template. That search eventually led me to Claude — and down an entirely different, incredibly visual rabbit hole of game development.</p><p>But that is a story for the next post.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=82843836054a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Following My Curiosity Into AI]]></title>
            <link>https://medium.com/@chrisglester/following-my-curiosity-into-ai-308d8df7e4ff?source=rss-6dbacd26aff2------2</link>
            <guid isPermaLink="false">https://medium.com/p/308d8df7e4ff</guid>
            <category><![CDATA[everyday-stories]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Chris Lester]]></dc:creator>
            <pubDate>Fri, 04 Sep 2026 12:32:01 GMT</pubDate>
            <atom:updated>2026-09-04T12:32:01.868Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*lm7iG03bF7I_Kz3h" /><figcaption>Photo by <a href="https://unsplash.com/@zanlazarevic?utm_source=medium&amp;utm_medium=referral">Zan Lazarevic</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>I’m not an expert in artificial intelligence. I’m not a developer. I’m not a researcher. I’m just someone who likes to understand how things work — whether that’s PC hardware, CAD modeling, game mechanics, or whatever new rabbit hole I fall into.</p><p>Recently, that rabbit hole has been AI.</p><p>I started exploring AI because I wanted to see what it could actually do for me. Not in a theoretical sense, but in a hands‑on, “can this help me build something?” way. And very quickly, I learned something important: AI is powerful, but it’s also imperfect, unpredictable, and occasionally hilarious.</p><p>This blog is where I’m documenting that journey — the good, the bad, and the blob‑shaped.</p><h3>How I Ended Up Here</h3><p>My first experiments were simple: could AI help me build an app? I tried one platform, hit limitations, jumped to YouTube, jumped to Google, jumped to other AIs. I even asked one AI how to use another AI. Sometimes it worked. Sometimes it failed spectacularly.</p><p>One of my earliest lessons was this:</p><blockquote><strong><em>AI is not perfect. Anything built by humans inherits human flaws.</em></strong></blockquote><p>Case in point: I once asked an AI to generate a coded 3D model for Tinkercad. It proudly produced a file. I imported it… and it was a blob. A literal blob. No shape. No structure. Just digital mush.</p><p>That’s when I realized:<br><strong>Always double‑check what AI gives you.</strong><br>Emails, code, models — all of it.</p><h3>The Rabbit Holes Get Deeper</h3><p>Eventually I found Claude, learned about its coding environment, and wondered if I could build something more complex. Then I stumbled onto a video about making games with AI.</p><p>That was it. The next rabbit hole.</p><p>I went back to Copilot and started designing a real‑time strategy game inspired by <em>Battle for Middle‑Earth II</em>. Together, we built a full world bible — factions, lore, mechanics, unit types, everything.</p><p>I dropped the whole thing into Claude and said, “Make an RTS based on this.”</p><p>Claude said: <em>Sure, no problem.</em></p><p>It generated a prototype.</p><p>I opened it.</p><p>And what I saw was… circles.<br>Circles everywhere.<br>A swarm of circle‑people marching across the screen, destroying my invisible fortresses.</p><p>That’s when I learned another lesson:</p><blockquote><strong><em>AI will not always fill in the blanks. If you aren’t precise, you get circles instead of elves.</em></strong></blockquote><h3>So What Am I Doing Here?</h3><p>I’m learning publicly.</p><p>I’m documenting:</p><ul><li>What I try</li><li>What works</li><li>What breaks</li><li>What surprises me</li><li>What I learn along the way</li></ul><p>I’m mostly here to explore — and share the process. But maybe I can also teach you something along the way.</p><p>I don’t know exactly where this journey leads. I don’t know what I’ll be building three months from now. But I do know this:</p><p>AI has already opened doors I didn’t expect.<br>It’s helped me build apps, generate models, and attempt a real‑time strategy game.<br>It’s shown me that curiosity is enough to get started.</p><h3>What You Can Expect From This Blog</h3><p>Honesty.<br>Curiosity.<br>Beginner mistakes.<br>Unexpected successes.<br>And a lot of learning.</p><p>I have many hobbies and interests, and this blog might wander through all of them. The cadence may change. The topics may shift. But the goal stays the same:</p><p><strong>Learn something. Build something. Share something.</strong></p><p>If you’re curious too, I hope you’ll join me.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=308d8df7e4ff" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>