<?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 James A. Manning on Medium]]></title>
        <description><![CDATA[Stories by James A. Manning on Medium]]></description>
        <link>https://medium.com/@thequickstartcreative?source=rss-13f43058462e------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*0EHOkz3qy1nAI8HSadNHNg.jpeg</url>
            <title>Stories by James A. Manning on Medium</title>
            <link>https://medium.com/@thequickstartcreative?source=rss-13f43058462e------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 00:09:38 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@thequickstartcreative/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[There is no F***ing thing as Building an App in Five Minutes]]></title>
            <link>https://medium.com/@thequickstartcreative/there-is-no-f-ing-thing-as-building-an-app-in-five-minutes-4725efbc9589?source=rss-13f43058462e------2</link>
            <guid isPermaLink="false">https://medium.com/p/4725efbc9589</guid>
            <dc:creator><![CDATA[James A. Manning]]></dc:creator>
            <pubDate>Sat, 10 Jan 2026 22:52:23 GMT</pubDate>
            <atom:updated>2026-01-10T22:52:23.541Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-4FEayEnOu2cIPapBlaIGw.png" /></figure><p>You’ve seen the videos. Some guy with a ring light and a shocked expression telling you he built Uber in five minutes. Or Airbnb. Or the next billion-dollar SaaS platform. The thumbnail shows a dashboard with numbers that would make a venture capitalist weep with joy. The comments are full of people asking where to sign up for the course.</p><p>Let me tell you what that guy actually built. He built the facade of Uber. He built the thing you see when you’re standing on the right side, at the right angle, in the right light. He built the Universal Studios backlot version of a Western town. Impressive as hell to walk through. Absolutely unfit for human habitation.</p><p>I’m not saying he’s lying about what he built. I’m saying he’s not telling you what he didn’t build. And that gap between what you see in the video and what it would take to actually ship that thing to real users, that gap is where your money goes to die.</p><p>There are three versions of any app. There’s the demo, the prototype, and production. Most of what you see on YouTube is the demo. Some of it, if you’re lucky, is a prototype. Almost none of it is production-ready.</p><p>The demo is the painted storefront. It works on the happy path. You type in the right address, request a ride, and boom, a car shows up on the map. Gorgeous. Impressive. Completely useless the moment anything goes wrong. Someone types an emoji in the address field and the whole thing crashes. That’s a demo.</p><p>The prototype is the house with the studs showing. You can see the wiring, the plumbing, the structure. It might survive some stress. You’ve thought about what happens when things go wrong. You’ve addressed the big blockers, the things that would make this catastrophically expensive or dangerous to run. You could release this to a small group of forgiving users who understand they’re testing something. It’s not production-ready, but it won’t explode in your face.</p><p>Production is move-in ready. Legacy system integration. Compliance. Security at scale. Support infrastructure when something breaks at 2am. All the engineering work that makes the difference between “it runs on my laptop” and “10,000 people can use this without my house burning down.”</p><p>The YouTuber showed you the backlot. He skipped the part where you need to make it structurally sound. He definitely skipped the part where you make it livable.</p><p>Every time you see one of these videos, ask one question: What would it take to ship this?</p><p>Not to make it look good in a screen recording. To actually ship it. To put it in the App Store. To charge money for it. To handle real users who will absolutely try to break it in ways you never imagined.</p><p>That question forces you to confront the invisible work. The security involved. The tech stack required to track drivers and users while protecting their privacy. The compliance requirements that change from state to state, city to city. The payment processing that has to work in 47 different jurisdictions with different tax laws. The support system for when someone gets charged twice or their ride never shows up.</p><p>They didn’t tell you about any of that. Not because they’re necessarily trying to scam you. Maybe they just don’t know. Maybe they’ve never actually shipped an app to the App Store. Maybe they built a web app once, which is different than getting something approved by Apple or Google and making actual money with it.</p><p>I’m listening to this kid on a podcast. He’s 19, maybe 20. Talking about his app that’s making $20,000 a month. Impressive, right? I’m half-paying attention until he explains what it does.</p><p>It searches for class action lawsuits with open settlements. You might qualify for money. Want to check? That’ll be $20 a month.</p><p>Wait. To <em>check</em>? Not to file. Not to get help with the process. Just to see a list of lawsuits you <em>might</em> be eligible for.</p><p>And there’s no freemium tier. No “search once for free.” You pay just to look. And once you’re in, good luck canceling.</p><p>The podcast drove curiosity. The app’s structure forces people into a payment plan they can’t easily escape. Somewhere in there he’s collecting twenty grand a month from people who probably don’t qualify for any of the settlements they’re looking at.</p><p>He didn’t build an app. He built a subscription trap. A roach motel where customers check in but can’t check out.</p><p>You want to know if someone actually knows what they’re talking about? Ask them nine questions. Three about resilience. Three about scale. Three about control.</p><p>Resilience means when something goes wrong, can it bounce back or does it crash? We understand this in our own lives. You lose a job, a marriage ends, a business fails. Can you bounce back? Same principle applies to software. Someone types garbage into a form field. Can your app handle that or does it fall over?</p><p>Scale means can it survive success? You planned a party for 20 people and 200 show up. If you thought that might happen, you have panel doors you can open, extra staff you can call, a system that lets this thing grow to 1,000 people without collapsing. If you didn’t plan for it, everyone’s standing in your kitchen and nothing works.</p><p>Control means are you protecting the data? Is authentication set up properly? The party was members only. You’re checking IDs at the door. You have security inside making sure nothing gets stolen and nobody gets hurt. Same thing with apps. Who can access what? How do you know they are who they say they are? What happens if someone tries to break in?</p><p>[Link to the nine questions]</p><p>If the person in the YouTube video can’t answer those questions, or if they dismiss them as something you’ll “handle later,” you’re looking at a demo. Not a prototype. Definitely not production. A demo.</p><p>Demos are fine. Demos have value. You can learn from a demo. But you need to know you’re looking at a demo. You need to understand that between what you see in that five-minute video and what you’d need to actually ship something real, there’s a Grand Canyon of work.</p><p>The YouTuber didn’t cross that canyon. He stood on one side, pointed to the other side, and said “looks easy enough.”</p><p>Don’t let those demos fool you. When you see someone build Uber in five minutes, pull up those nine questions. See if any of them were addressed or even acknowledged. Until all nine are answered, it’s a demo. They built a facade. It’s cool as hell. It’s not fit for humans to actually use.</p><p>The YouTuber stood on one side of the canyon and said it looks easy enough. You want to actually cross it? Start small. Build a tool that solves one problem you actually have. Not Uber. Not Airbnb. A script that automates the thing you do manually every Monday. A dashboard that pulls three numbers you check constantly.</p><p>Build something boring that works. Then build another one. That’s your toolbox.</p><p>The painted storefronts can stay on YouTube. You’re building actual houses.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4725efbc9589" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Automation Isn’t a Workforce Strategy]]></title>
            <link>https://medium.com/@thequickstartcreative/automation-isnt-a-workforce-strategy-a2dca1fcc8e6?source=rss-13f43058462e------2</link>
            <guid isPermaLink="false">https://medium.com/p/a2dca1fcc8e6</guid>
            <dc:creator><![CDATA[James A. Manning]]></dc:creator>
            <pubDate>Mon, 29 Dec 2025 20:16:55 GMT</pubDate>
            <atom:updated>2025-12-29T20:16:55.403Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yJingoljklnv5wmEu26deA.jpeg" /></figure><p><strong>Why Robot Training Should Be the Beginning, Not the End</strong></p><p>At the end of almost every basic robot operation class I teach, someone pulls me aside and asks a simple question:</p><p>“So… what’s the next class?”</p><p>That question matters more than most people realize.</p><p>It tells me two things immediately. First, the person in front of me is curious. Second, they’ve sensed something important on their own: learning the basics is not the same as becoming capable. And somewhere between those two ideas is an opportunity most manufacturers are missing.</p><p>I spend a lot of time with companies investing in automation because of labor challenges. Retirements. Hiring gaps. Inconsistent throughput. The reasons are real and valid. Automation helps. But automation does not replace the need for a workforce development strategy. In many cases, it actually makes that need more urgent.</p><p>The problem is that many companies don’t have a workforce development strategy at all. What they have is training. And they’re not the same thing.</p><h3>Training Solves Today’s Problem</h3><p>Workforce Development Solves Tomorrow’s</p><p>Training is transactional. You train someone so they can operate a system, pass a certification, or complete a task safely. It answers the question: <em>Can this person do the job right now?</em></p><p>Workforce development answers a different question: <em>How does someone grow here?</em></p><p>That distinction gets blurry in manufacturing, especially in smaller shops. Robot training is often viewed as a necessary expense tied to equipment purchases. You buy a robot. You train an operator. Box checked. Move on.</p><p>But when someone asks what’s next, they’re telling you they don’t want to stay at the floor-level forever. They’re asking, sometimes without realizing it, whether there is a path forward. More responsibility. More trust. More pay. More agency.</p><p>If the only answer is “There isn’t one,” that curiosity eventually turns into disengagement. Or attrition.</p><h3>Automation Changes Roles, Not Just Processes</h3><p>One of the quiet truths about automation is that it shifts where expertise lives.</p><p>Robots remove certain manual tasks, but they create new decision points upstream and downstream. Someone has to understand process limits. Someone has to troubleshoot intelligently. Someone has to know when a problem is mechanical, procedural, or programming-related. Someone has to own changeovers, quality drift, and recovery.</p><p>Those responsibilities don’t appear automatically just because a robot is installed. They require people who grow into them.</p><p>Without intentional pathways, companies end up with a small handful of “go-to” experts and a wide base of operators who never get the chance to move beyond button-pushing. That’s risky. It’s also exhausting for the few experts carrying all the weight.</p><h3>Why Small Manufacturers Feel Stuck</h3><p>I don’t think this is negligence. I think it’s structural.</p><p>Many small manufacturers don’t have relationships with local high schools, trade programs, or community colleges. They don’t run summer job programs. They don’t have defined entry points for young people to experience the environment before committing to it.</p><p>Inside the company, roles are often informal. People wear multiple hats. Growth happens ad hoc. Someone “moves up” because they’ve been around long enough, not because there’s a clear capability ladder.</p><p>When automation enters that environment, it exposes the gaps. Suddenly you need deeper technical understanding, but you haven’t built a way to grow it internally. Training becomes episodic. Workforce development remains undefined.</p><h3>What a Workforce Development Strategy Actually Looks Like</h3><p>A workforce development strategy doesn’t need to be complicated. But it does need to be explicit.</p><p>At its core, it answers a few questions clearly:</p><p>What are the capability levels in our operation, from entry-level to advanced?<br> What does someone need to know and be able to do at each level?<br> How does responsibility expand as capability increases?<br> How does compensation reflect that growth?</p><p>In an automated environment, that often means recognizing that not everyone needs to become a programmer, but some people should be able to progress from operator to advanced operator, to technician, to specialist.</p><p>Training supports those steps. It doesn’t define them.</p><p>A strategy also extends beyond your walls. Partnerships with local schools. Internships. Summer roles. Guest lectures. Plant tours. These aren’t charity. They are long-term investments in a talent pipeline that understands your reality before day one.</p><h3>The Missed Opportunity in Robot Training</h3><p>When someone finishes a basic robot course and asks what’s next, they’re standing at a fork in the road.</p><p>One path says: “This is as far as it goes.”<br> The other says: “Here’s how you build expertise.”</p><p>Companies that choose the second path don’t just get better operators. They get people who understand the system, take ownership, and stay longer because they can see a future.</p><p>Automation amplifies whatever system you already have. If your system only supports short-term training, you’ll feel that limitation quickly. If it supports long-term development, automation becomes a force multiplier for both productivity and people.</p><p>That’s the difference between buying robots and building capability.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a2dca1fcc8e6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What Actually Matters When You Build Apps (2025 Recap)]]></title>
            <link>https://medium.com/@thequickstartcreative/what-actually-mattered-when-you-build-apps-2025-recap-7919763d5fa9?source=rss-13f43058462e------2</link>
            <guid isPermaLink="false">https://medium.com/p/7919763d5fa9</guid>
            <dc:creator><![CDATA[James A. Manning]]></dc:creator>
            <pubDate>Sat, 27 Dec 2025 23:13:50 GMT</pubDate>
            <atom:updated>2025-12-28T01:00:37.972Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0BQbHQofJ1-5dzYVTkfVqw.png" /></figure><p>2025 was supposed to be the “Year of the AI Agent.” For developers, that was somewhat accurate. Claude Code, OpenAI Codex, AntiGravity, and a host of other agentic IDEs made real improvements. But for most people, this was the year of building. And it turns out you can build some amazing things.</p><p>I had a lot of fun with AI this year. But rather than talk about the tools, I want to talk about the lessons I learned. These lessons are tool-agnostic. And interestingly enough, a lot of them align with what I teach my students when they’re learning to program welding robots.</p><p>That leads me to the first lesson.</p><ol><li><strong>Being Prepared Is the First Job</strong></li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ekcl8D3WMg1femsP5bTbZw.png" /></figure><p>When I teach robot programming, the final lab is never about programming the part.</p><p>That surprises people. They assume the job is writing the program, moving the robot, making sparks. But by the time we get there, most of the real work should already be done.</p><p>The last lab is about preparation. We power up the pendant and don’t touch the program yet. We find the parameters. We record the parameters. We set the defaults. We make sure the robot is in the right mode, confirm the tooling, check the coordinate systems, verify the job setup. Only after all of that do we write the program.</p><p>I’ve watched what happens when someone skips those steps. They get impatient, jump straight to jogging the robot around, and within fifteen minutes they’re staring at error codes they don’t understand. The program didn’t fail. The preparation failed. And now they’re spending twice as long troubleshooting something that never should have happened in the first place.</p><p>Building software works the same way. Preparation is not documentation. Documentation comes later, after you’ve learned something. Preparation is understanding the job so deeply that the tooling becomes almost incidental. You should be able to explain the entire process, start to finish, without mentioning AI, models, or frameworks at all.</p><p>Who is this for? What are they actually doing today? Where does it break? What happens when it goes wrong? What does “good” look like in the real world, not on a whiteboard?</p><p>If you can’t answer those questions calmly and clearly, you’re not ready to build yet. That’s not a criticism. That’s just the order of operations.</p><p><strong>2. Models Are Secondary to Architecture</strong></p><p>I think about this every time I see someone build an app around a specific model’s quirks. They optimize for GPT-4’s behavior, or they tune everything around Claude’s tendencies, and then six months later the model updates or the pricing changes and suddenly they’re rewriting half their system.</p><p>Models matter, of course. Some are conservative. Some are confident to a fault. Some infer aggressively when you’d rather they asked for clarification. You need to know those differences. But the model is not the system. The architecture is.</p><p>If your app collapses the moment you change models, that’s not a model problem. That’s a design problem. The model should sit behind clear interfaces. Inputs and outputs should be explicit. Retrieval, validation, and post-processing should happen independently of whatever model you’re calling.</p><p>When you build it that way, the model becomes a replaceable component. And replaceability is what gives you stability over time.</p><p><strong>3. Go Deep on a Stack, Not Broad Across Tools</strong></p><p>When you know your stack deeply, you stop reacting to problems and start anticipating them. You know which warnings matter and which ones you can safely ignore. You know where things tend to break under load. You know which shortcuts are safe and which ones will cost you later.</p><p>I’ve used the same basic stack across almost every project I’ve shipped this year. React, TypeScript, Supabase, and whichever AI APIs the project needs. Not because I think those are the best tools in some objective sense, but because I know them well enough to build calmly under pressure. When something goes wrong at ten at night, I’m not frantically Googling basic syntax. I’m focused on the actual problem.</p><p>Chasing new tools feels productive. Every week there’s some new framework that promises to make everything easier. But familiarity is what actually lets you move quickly without creating new problems in the process.</p><p><strong>4. Most of Us Are Not Developers. Build Like It</strong></p><p>The systems I build are intentionally small. They don’t integrate with enterprise infrastructure. They don’t handle payments. They don’t collect sensitive personal data. Most of them live on simple hosting and serve small teams with defined needs.</p><p>That’s not playing small. That’s recognizing where my lane ends.</p><p>Most builders today are not enterprise developers, and that’s completely fine. The mistake is pretending otherwise. Build small. Build scoped. Avoid the things you can’t properly support. When you reach the edge of that lane, when you need compliance or scale or payments or heavy integration, discipline means handing it off to someone who actually does that work professionally.</p><p>There’s no shame in that. Knowing where your expertise ends is part of building responsibly.</p><p><strong>5. Real World Problem Solving Beats Demos</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XL_Ape34_veMPNARqgv47w.png" /></figure><p>I’ve never seen a demo save a production floor.</p><p>Demos are for showing what’s possible. They’re impressive for about fifteen minutes, and then someone asks, “Okay, but can it do this thing we actually need?” and the answer is usually no, or not yet, or we’d have to rebuild most of it.</p><p>Real systems are for showing up every day and not breaking. The projects that last always start in the same place. Someone is wasting time. Someone is repeating a task they hate. Someone is compensating for a broken process with manual effort because nothing else exists to help them.</p><p>You don’t automate chaos. You clean it up first. You understand why the process is painful before you try to make it faster. Otherwise, you just get faster failure with better graphics.</p><p><strong>6. Document Everything, Deployed or Not</strong></p><p>Every project I work on gets four files, whether it ships or not. Architecture, which describes how the system is structured and why. Frontend, which captures the user-facing decisions and patterns. Backend, which documents the data flow and integrations. And Resolutions.</p><p>That last one matters most. The resolutions file is where I write down what broke, why it broke, and how I fixed it while it’s still fresh. Not a polished retrospective. Just honest notes about what went wrong and what actually solved it.</p><p>It’s root cause analysis applied to building software. Over time, that file becomes institutional memory. I can look back at a project from six months ago and remember not just what I built, but what I learned while building it. What traps I fell into. What I’d do differently.</p><p>Systems don’t improve because they succeed. They improve because failure is captured instead of buried. The resolutions file is how I make sure I don’t solve the same problem twice without realizing it.</p><p><strong>7. Trust Beats Capability</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fYM5pk_nIauvbdkNC8KPcw.png" /></figure><p>When I built the chatbot for industrial automation support, I spent more time on the refusal logic than on the generation logic. The system uses a confidence check before it answers anything. If it’s not confident enough that it understood the question correctly, it doesn’t guess. It tells the user it can’t help with that specific question and suggests they contact a human.</p><p>That might sound like a failure. Forty percent of questions don’t get answered by the AI. But the alternative was worse. In an industrial setting, a wrong answer isn’t just a bad user experience. It’s a potential safety issue. Someone might configure equipment incorrectly based on a hallucinated response.</p><p>Users don’t want systems that guess confidently. They want systems that know their limits and behave predictably when they reach them. A system that says “I don’t know” when it genuinely doesn’t know builds more trust over time than one that always produces an answer.</p><p>Constraints are not limitations. They’re trust mechanisms. And trust is what turns a prototype into something people actually rely on.</p><p><strong>8. The Ceiling for Non-Technical Builders Has Collapsed</strong></p><p>This part really is different now. I’m not saying that to hype the tools. I’m saying it because I’ve watched the gap close in real time.</p><p>For years, non-technical people could prototype ideas but not finish them. You could mock something up in a no-code tool, demonstrate the concept, get people excited. And then you’d hit a wall. Deployment. Authentication. Storage. Backend logic. The handoff to “real developers” was where momentum went to die. You’d wait weeks. You’d explain your vision over and over. And by the time it finally shipped, if it shipped at all, it didn’t quite match what you’d imagined.</p><p>That ceiling has collapsed for a certain class of work.</p><p>Think about what couldn’t happen a year ago. You couldn’t generate an infographic end-to-end without handing the final polish to a designer. You couldn’t build a working backend without understanding infrastructure. Video couldn’t produce coherent audio. Images couldn’t handle text reliably.</p><p>Now the person who understands the process can often take it all the way to production, as long as they stay disciplined about scope. That doesn’t mean everything is possible. Enterprise integration, compliance, payments, scale. Those still require real engineering. But for focused tools serving small teams with defined problems? The handoff is no longer mandatory.</p><p>The advantage has shifted to the person who understands the job. That’s a fundamental change in who gets to finish things.</p><p><strong>9. Prompting Still Matters, Just Not How People Think</strong></p><p>There’s a misconception that better models make prompting less important. The opposite is true.</p><p>Prompting isn’t about finding clever wording or magic phrases that unlock hidden capabilities anymore. It’s about constraint design. You define the model’s role explicitly. You specify what output formats are acceptable. You establish how the system should handle uncertainty, what it should do when the context isn’t sufficient to proceed.</p><p>Better models raise the stakes on this. A powerful model with vague instructions is harder to control, not easier. It has more room to interpret, more freedom to infer, more ways to drift from what you actually wanted.</p><p>I’ve seen this firsthand with different models. Gemini tends to be confident even when it shouldn’t be. If it doesn’t know something, it’ll often give you an answer anyway, and occasionally that answer will include useful inferences you didn’t expect. GPT tends to be more conservative, more likely to caveat or refuse. Neither behavior is right or wrong in the abstract. But if you don’t account for those tendencies in your prompts, you’ll get surprised by results that don’t match your expectations.</p><p>Prompting is about making the model’s behavior predictable. And predictability requires deliberate constraint, not more sophisticated phrasing.</p><p><strong>10. Memory, Data, and Security Are the Final Chasm</strong></p><p>Most projects don’t fail because of models. They fail because of state.</p><p>What gets remembered. What gets overwritten. Who has access to what. Where data actually lives.</p><p>Memory is the first trap. Everyone talks about context windows getting larger, about RAG systems that can pull in relevant documents, about vector stores that enable semantic search. And all of that helps. But none of it solves the fundamental problem of deciding what your system should know at any given moment. Larger context windows don’t tell you what to put in them. RAG doesn’t clean messy data. If the underlying information is wrong or inconsistent or outdated, the system will faithfully reproduce the problem.</p><p>I built a document processor specifically because I couldn’t use most of my source material in its original form. Technical documentation written for one audience, in one format, structured in ways that made sense for manuals but not for AI retrieval. The RAG system couldn’t work until the data was clean. That preparation step took longer than building the actual chatbot.</p><p>Data quality is the silent prerequisite. Nobody wants to talk about it because it’s boring, but it determines whether anything downstream actually works.</p><p>And then there’s security. This is where I see the most reckless behavior.</p><p>I watch people on YouTube demonstrate lead generation apps for medical facilities without mentioning HIPAA once. Builders in the UK talking about apps for US companies without any discussion of data residency requirements. People handling personal information without understanding what SOC 2 means or why it matters.</p><p>That’s not innovation. That’s liability waiting to happen.</p><p>If you don’t understand where data lives, who can access it, what regulations apply, and what happens when something goes wrong, you’re not building a product. You’re building a problem that hasn’t exploded yet.</p><p>These three things, memory, data, and security, are where projects quietly collapse. They don’t make good demos. They don’t generate excitement. But they determine whether your system survives contact with reality.</p><p><strong>Closing Thought</strong></p><p>What’s changed in 2025 isn’t that AI suddenly became intelligent.<br> It’s that the tools finally stopped getting in the way of good judgment.</p><p>For the first time, the person who understands the work can carry an idea all the way to something real without waiting for permission. That’s a meaningful shift. It changes who gets to finish things. It changes which problems get solved. And it raises the bar on responsibility at the same time.</p><p>Because when you can ship, you don’t get to hide behind demos anymore. You don’t get to blame tooling. You don’t get to say, “That’s someone else’s job.” The decisions are yours. What you automate. What you constrain. What you refuse to build. What you hand off.</p><p>The next phase isn’t about smarter models or bigger context windows. It’s about builders developing taste. Judgment. Restraint. The ability to look at a powerful system and say, “Not yet,” or “Not like that,” or “This is enough.”</p><p>That’s where the advantage is going to come from.<br> Not cleverness. Not velocity. Discipline.</p><p>The people who win won’t be the ones who build the most.<br> They’ll be the ones who know what <em>not</em> to build, and why.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7919763d5fa9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[2025 Was Supposed to Be the Year of the Agent]]></title>
            <link>https://medium.com/@thequickstartcreative/2025-was-supposed-to-be-the-year-of-the-agent-7ffb4c0bc596?source=rss-13f43058462e------2</link>
            <guid isPermaLink="false">https://medium.com/p/7ffb4c0bc596</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-and-creativity]]></category>
            <category><![CDATA[creativity]]></category>
            <dc:creator><![CDATA[James A. Manning]]></dc:creator>
            <pubDate>Sun, 14 Dec 2025 15:24:00 GMT</pubDate>
            <atom:updated>2025-12-14T15:24:00.760Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*iaQISVP4gGOqYvbJmghUoA.jpeg" /></figure><p>Agents were going to replace software engineers. Workflows were going to design and deploy themselves. AGI was always just around the corner. If you believed the hype, this was the year human effort finally became optional.</p><p>That&#39;s not what happened.</p><p>What actually happened is that the industry finally paved the last mile to production. And for the first time, the tools got good enough to show us where the real work has always lived.</p><p>Nano Banana Pro made image generation reliable enough to move past hobbyist experiments. Loveable integrated Supabase and collapsed the gap between prototyping and production. Google Build opened Firebase and Google Cloud to people who&#39;d never touched IAM. Meta released SAM 2 and showed what segmentation looks like when research catches up to real problems. n8n added AI agent that designed workflows from a prompt.</p><p>AI video crossed its own threshold - you could finally generate finished video with sound, voice, and timing without stitching together half a dozen tools.</p><p>Before 2025, you could design a flyer, but you couldn&#39;t ship the final asset. You could generate images, but not a finished campaign. You could get video clips, but adding dialogue and pacing meant manual work and external tools. You could vibe-code an interface, but the backend left most projects stuck in prototype limbo.</p><blockquote>This is what 2025 actually solved. Not creativity. Not intelligence. Production.</blockquote><p>You could always build the prototype and prove the concept. What you couldn&#39;t do - not without being a software engineer or hiring one - was ship it. Deploy it. Make it work for someone other than yourself.</p><p>The road to production was broken. And once that road got paved, something else became obvious.</p><p>Look at what happened with the YouTube automation channels. They started the year teaching n8n workflows - how to connect these nodes, how to set up that integration, how to wire together the technical pieces. Pure execution tutorials. Tool-focused content.</p><p>By the end of the year, they&#39;d completely flipped the message. Same creators, different take: stop obsessing over the tools. Don&#39;t burn all your time learning automation syntax. Think about systems. Solve actual problems. Build things that matter.</p><p>That’s not an accident; that’s what happens when the production barrier drops.</p><p>When n8n was the main game, you had to be technical enough to wire everything together manually. The bottleneck was infrastructure. But when Loveable and Google Build showed up and gave you production-ready backends, the constraint shifted. It stopped being &quot;can you configure the deployment&quot; and became &quot;do you actually understand what you&#39;re building?&quot;</p><p>The tools got out of the way. What&#39;s left standing is the thing that was always there: **judgment, taste, domain expertise, and the ability to think in systems.**</p><p>When everyone has access to the same path to production, the gap isn&#39;t technical anymore. It&#39;s human. Do you understand the problem you&#39;re solving? Do you know what&#39;s worth building? Can you see the system, not just the feature?</p><p>Before 2025, you could hide behind the technical gap. &quot;I would&#39;ve shipped it if I knew how to set up a database.&quot; &quot;I would&#39;ve deployed it if I could figure out the backend.&quot; The broken road gave you an excuse.</p><p>Not anymore.</p><p>**2025 didn&#39;t make humans optional. It made human expertise non-negotiable.**</p><p>The limiting factor isn&#39;t access to production infrastructure anymore. It&#39;s whether you know what you&#39;re building and why. Whether you have the craft to make decisions about structure, behavior, and restraint. Whether you understand your domain well enough to spot the difference between a demo and something that actually solves a problem.</p><p>The YouTubers who flipped their message got this. The production tools didn&#39;t make their expertise less valuable - they made it *more* valuable. Because now the people watching could actually ship what they envisioned. But only if they had something worth shipping in the first place.</p><p>If you&#39;re a builder - if you have craft, if you understand your domain, if you can think in systems - the tools are finally good enough. The last mile is paved.</p><p>What’s left is you. And it’s important that you understand this.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7ffb4c0bc596" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Is Learning & Development Is Making the Same Mistake That Cost Canon 5 Years]]></title>
            <link>https://medium.com/@thequickstartcreative/is-learning-development-is-making-the-same-mistake-that-cost-canon-5-years-8bd2c3c4457e?source=rss-13f43058462e------2</link>
            <guid isPermaLink="false">https://medium.com/p/8bd2c3c4457e</guid>
            <category><![CDATA[lds]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[training]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[James A. Manning]]></dc:creator>
            <pubDate>Sun, 09 Nov 2025 20:19:07 GMT</pubDate>
            <atom:updated>2025-11-14T04:06:57.770Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yaelTVKbM43jiy3uhh0pSg.png" /></figure><p>The photography industry used to be certain about what counted as a professional camera. A real camera had weight. Presence. A mirror that snapped loud enough to make you feel like serious work was happening. Canon and Nikon defined that world. They owned most of the interchangeable-lens camera market and an even larger share of the professional space. For decades they didn’t just sell gear. They set the standard for what “professional” meant.</p><p>Then, in October 2013, Sony released the a7. A full-frame mirrorless camera that broke every visual cue the industry relied on. It was smaller. Lighter. Quieter. Next to a Canon 5D Mark III, it barely looked real. And that’s how most professionals reacted. They dismissed it as a toy.</p><p>What Sony had actually introduced was a question. What happens when the thing we treat as essential — in this case, the mirror — isn’t necessary anymore?</p><p>Canon and Nikon didn’t answer that question for five years. By the time they did, Sony had already spent half a decade exploring what a camera could become.</p><p>I keep coming back to that story because I’m starting to wonder if L&amp;D is standing in the same moment. I’m not saying the industry is repeating the mistake. I’m saying the conditions feel similar enough that it’s worth asking whether we’re headed down that path.</p><p>When I talk with L&amp;D teams about AI, the conversation often splits into opposites. Some are anxious about being replaced. Others make big declarations about transformation. But very few people are doing the uncomfortable middle work of building systems that rethink how learning happens altogether. That’s where the interesting questions live. And those questions sound different from what we usually ask.</p><p>Instead of “How do I use AI to make my Storyline course better?” the more revealing question is “Should a Storyline course even be the default solution anymore?”</p><p>The first is optimization. The second is curiosity. And curiosity is the part that matters.</p><p>Canon and Nikon spent years optimizing DSLRs while Sony was asking the more disruptive question. What happens if you get rid of the mirror entirely?</p><p>So when I see L&amp;D asking the first type of question more often than the second, I can’t help but wonder if we’re sticking close to what we know instead of exploring what might be possible.</p><p>We’ve seen something similar before. When xAPI came along, it offered a chance to understand learning more deeply. Better data. Better insight. A more realistic picture of how people learn in real environments. It asked us to stretch ourselves technically. Not dramatically. Just enough to evolve.</p><p>Most of the field held back. It wasn’t hostility. It was hesitation. It was waiting for tools to mature, for vendors to catch up, for best practices to settle. Those are reasonable instincts. But they also slowed adoption to a crawl.</p><p>Now, with AI, the same hesitations are showing up. People want certainty before they experiment. They want executive buy-in before they explore. They want to understand the entire landscape before they build anything. Those instincts make sense. But they’re also the same instincts Canon and Nikon had in 2013.</p><p>This is why I’m not handing out verdicts. I’m asking a different question. Are we seeing the early signs of another moment where the field sticks with the familiar while someone else pushes forward?</p><p>My work at Miller shapes how I think about this. When I teach welders to program robots, the breakthrough isn’t the new skill. It’s the moment when they realize their existing expertise transfers into the new environment. Instructional designers already have the methodology. Needs analysis works. Learning theory holds up. ADDIE makes sense. None of that is obsolete.</p><p>What’s changing is the range of possible solutions.</p><p>For decades, “build a course” was the default answer. Today there’s a whole ecosystem of options. Micro-apps, conversational agents, AR overlays, flow-of-work guidance, AI tutors, analytics dashboards that flag performance gaps in real time. The question isn’t whether courses disappear. The question is where they fit when the solution space is bigger than ever.</p><p>And that’s the part I’m trying to shine a light on. Not a conclusion. A curiosity.</p><p>Are we exploring that expanded space? Or are we trying to fit new problems into old containers because those containers are familiar?</p><p>Even the anxiety around technical skills fits this pattern. Whenever I say instructional designers need basic literacy in markdown, JSON, JavaScript concepts, and APIs, some folks tense up. But those skills aren’t about becoming developers. They’re about understanding the shape of the new solution space. They’re about having enough familiarity to explore. Enough comfort to try things. Enough knowledge to recognize what’s possible.</p><p>You don’t learn these skills in a classroom. You learn them by building things. Trying things. Asking “what if?” And that experimentation doesn’t require organizational transformation. It just requires permission. Sometimes self-permission is the hardest part.</p><p>Because here’s the real question hanging over all of this. When a new technology shifts the landscape, do we explore early while everything is still malleable? Or do we wait until the rules are written and the standards are fixed?</p><p>Sony didn’t win because they had better answers. They won because they were willing to explore while everyone else waited.</p><p>So I’m not saying L&amp;D is repeating the Canon and Nikon mistake. I’m asking whether we might be drifting in that direction without noticing. And if we are, now is the moment to look up.</p><p>The shift we’re going through doesn’t erase learning theory or design craft. It expands the size of the canvas. That expansion asks three things from us: curate context that isn’t in public LLMs, architect learning systems that support performance as it happens, and pay attention to the patterns in what people struggle with.</p><p>We already have the expertise to do that. We just have to give ourselves permission to explore.</p><p>Because the biggest risk isn’t that AI changes everything. The biggest risk is that we treat it as a small tool for tweaking the old way instead of asking whether the old way needs rethinking.</p><p>That’s the question I’m sitting with. That’s the question this moment reminds me of. And I think it’s one the field deserves to ask together.</p><p><em>James works as a trainer at Miller Electric, teaching people to program welding robots and developing automation education solutions. He writes about finding human pace in an age of acceleration and builds agentic learning systems that bridge corporate L&amp;D expertise with manufacturing training needs. He believes the question isn’t whether your job will be automated, but which parts of your work are applying expertise to objects versus applying expertise toward people — and that distinction makes all the difference.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8bd2c3c4457e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Affinity and the Trap of Building Against a Giant Like Adobe]]></title>
            <link>https://medium.com/@thequickstartcreative/affinity-and-the-trap-of-building-against-a-giant-5d1d58aafeaf?source=rss-13f43058462e------2</link>
            <guid isPermaLink="false">https://medium.com/p/5d1d58aafeaf</guid>
            <dc:creator><![CDATA[James A. Manning]]></dc:creator>
            <pubDate>Mon, 03 Nov 2025 03:30:01 GMT</pubDate>
            <atom:updated>2025-11-04T04:09:16.532Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*oVzmAsLvkQZDfcXu.png" /></figure><p>Affinity did not collapse. They did not run out of talent or vision. They built beautiful software that millions of designers chose willingly. They carved out a space in a market most believed was unwinnable. That deserves respect.</p><p>But their story offers a lesson worth sitting with. It’s about positioning, and how the story that launches you can quietly limit what you become.</p><p>When Affinity emerged, the pitch landed instantly: professional-grade creative tools, no subscriptions, fast, elegant, and independent. It felt like a breath of fresh air in a world tired of renting the tools we rely on to make a living. There was pride in switching. People didn’t just buy Affinity — they defected to it.</p><p>It wasn’t just a product choice. It was a declaration. <br>Control. Ownership. Autonomy. A creative tool that felt like it belonged to <em>you</em> again.</p><p>That kind of energy is powerful in the beginning. It builds community quickly and creates loyalty that money can’t buy. But markets are not powered by sentiment. They move according to behavior, and behavior evolves.</p><p>And when your identity is rooted in standing against something, the very thing you resist becomes the anchor tied to your ankle.</p><p>History has seen this play out before. <br>In the early 1960s, Avis launched one of the most famous campaigns in advertising history: <em>“We’re number two, so we try harder.”</em> Smart, sharp, self-aware. It worked — and it also locked them into a frame. Every win, every loss, every move was measured against Hertz. They lived in comparison, even in victory.</p><p>Apple took a different path. They also started with rebellion. “1984.” “I’m a Mac.” Anti-big-tech energy before big tech was a category. Yet Apple’s power didn’t come from fighting IBM or Microsoft. The rebellion was the doorway, not the destination. Their true identity was built on imagination and possibility. “Think Different” wasn’t an argument. It was an invitation.</p><p>Affinity’s origin looked more like Avis than Apple. It rallied those who were tired of Adobe’s grip. It built a home for people who wanted to make things without feeling like a tenant in their own studio. That matters. For a long time, it was enough.</p><p>But rebellion sparks revolutions. It does not sustain eras.</p><p>While Affinity perfected the craft-driven desktop studio, design practice itself shifted. Designers became multi-disciplinary, rapid-cycle creators. They iterate inside collaborative loops. They build systems that live across platforms and channels. The job expanded from making beautiful things to or chestrating entire brand universes that evolve in real time.</p><p>Figma didn’t try to be a better Photoshop. It didn’t fight Adobe. It stepped around Adobe entirely. It imagined design not as mastery in isolation, but as shared momentum. Real-time work. Connected thinking. Creation as a team sport, not a solitary craft.</p><p>Affinity’s strength — focus, precision, personal control — remained rooted in the earlier era of design practice. They refined what once defined excellence, at the same moment the meaning of excellence changed.</p><p>When a brand’s center of gravity is rejection, evolution becomes tricky. You can update features, ship new tools, optimize performance — but your story stays tethered to the thing you walked away from. You become a refuge for those who want the past preserved, not a vehicle for those charting the next frontier.</p><p>Sanctuaries are safe. They are also still. Sometimes stillness becomes stasis.</p><p>This is why Canva acquiring Affinity is not a failure or surrender. It’s a pivot point. Canva sees in Affinity the engine. The craftsmanship. The purity. But Canva also sees the future of creative work: workflow, distribution, acceleration, shared execution, AI as assistant not overlord, and the reality that content is now a stream, not an export.</p><p>Affinity built the studio. Canva built the highway. Now we see whether they merge into something new.</p><p>This isn’t really about creative tools. It’s a pattern in product strategy, and it applies far beyond design software.</p><p>You can build momentum by positioning against a dominant force. It feels righteous. It feels bold. It builds fast loyalty, especially in communities that feel underserved.</p><p>But anchoring your identity in opposition is a short-term accelerant, not a long-term compass. Anger fuels escape, not direction. The companies that endure don’t merely run from the past; they name a future and walk toward it with clarity.</p><p>Affinity succeeded at independence. Now the question becomes: can they succeed at evolution?</p><p>Are you creating an escape from the present, or a path toward what comes next?</p><p>One burns hot. The other: long and bright.</p><p>Affinity gave us a spark. The next movement in creative tools belongs to whoever turns that spark into a horizon — not by defining who they stand against, but by defining what they stand to build.</p><p><em>Originally published at </em><a href="https://mynameisjames.substack.com/p/affinity-and-the-trap-of-building"><em>https://mynameisjames.substack.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5d1d58aafeaf" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Darien Gap of Vibe Coding]]></title>
            <link>https://medium.com/@thequickstartcreative/the-darien-gap-of-vibe-coding-7e8b6d0b8da6?source=rss-13f43058462e------2</link>
            <guid isPermaLink="false">https://medium.com/p/7e8b6d0b8da6</guid>
            <dc:creator><![CDATA[James A. Manning]]></dc:creator>
            <pubDate>Fri, 31 Oct 2025 00:37:29 GMT</pubDate>
            <atom:updated>2025-10-31T00:37:29.978Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FBaQ88cs3XHxa5JWtgOHXw.png" /></figure><blockquote>The Darien Gap is about 60 miles of dense rainforest, swamps, fast rivers, and rough terrain between Panama and Colombia. There are no continuous roads through it, which makes crossing dangerous and logistically brutal. I use it as a metaphor for the backend work that turns a pretty prototype into a real product.</blockquote><p>Think of vibe coding like walking to the edge of a continent. Gemini gets you to the shoreline fast. It hands you a boat, a compass, and a killer view. Then you look out and there it is, the Darien Gap, a brutal stretch nobody wants to film. That gap is the backend reality. It is authentication, databases, CI, permissions, environment variables, cloud consoles, scaling, and security. It is where demos stop being fun and start being real work. People love showing you the shoreline because it looks great on camera. The part that kills projects is the crossing, and too many creators pretend there is an easy ferry. There is not.</p><p>This is not a political post. If you want to make this about immigration, fuck off. We are talking tooling, expectations, and the technical choke points that turn a prototype into something you can actually ship.</p><h3>The truth nobody wants in a 30-minute demo</h3><p>You’ve seen the clips: build an app in 30 minutes, no-code, no problem. Thumbnails sell clicks. Videos sell hope. The truth is more boring and more valuable. The UI is the easy part. The frontend is what looks sexy in a timelapse. It does not prove your product can survive real users, real data, or real attacks.</p><p>Gemini gets you 0 to 90 faster than anything I’ve used. It is brilliant at ideation, layout, and moving from idea to working prototype in hours instead of weeks. That speed is a game changer. But when you try to wire Firebase or other Google services for auth and storage, you land in Google Cloud Console. If you have never lived in that world, it feels like being dropped into a professional kitchen mid-service. Permissions, APIs, service accounts, IAM roles — it is a pro environment built for operators, not demo makers. That is not a design flaw in Gemini. It is a scope decision. Gemini focuses on frontend velocity. It is not trying to replace an ops team.</p><h3>My workflow, plain and simple</h3><p>I prototype and iterate in Gemini until I have a clear PRD. I ask Gemini to produce the roadmap, the likely pitfalls, and sample code snippets. That road map becomes the handoff.</p><p>Then I move to Loveable. Loveable is the bridge across the gap. Supabase is integrated. GitHub is connected. One prompt can spin up the backend, wire the database, and keep version control sane. Instead of juggling Google consoles, I get an environment that maps to real deployment practices without forcing me to become a cloud admin.</p><p>Because I do the thinking first, my iteration in Loveable drops from roughly 40 prompts to about 15. Because I usually work on these projects part-time, a real production-ready app ends up taking about 60 focused hours. If I worked on it full-time, I could ship in a week. Sixty hours is not a marketing sound bite. It is an honest metric that matters.</p><h3>What content creators should stop doing and what they should start doing</h3><p>I am not here to dunk on creators. Many of them lower the fear barrier and get people started. That is valuable. But the narrative needs one tiny addition: transparency.</p><p>If you make a 30-minute app video, say this at the end: “This is the frontend. I preconfigured the backend for this demo. I’ll show the auth and infra setup in a follow-up.” That single sentence changes outcomes for beginners. It sets expectations and points people toward the crossing instead of pretending the shoreline is the whole trip.</p><h3>Real, practical advice</h3><p>Use Gemini to move fast and think. Get the UI, the flows, and a solid PRD. Treat the PRD like a contract with reality. Include data models, auth flows, and error states. Move the project to a tool with built-in infra for the final mile. Supabase plus GitHub integration is non-negotiable if you want sane deployments. Run audits with Claude or ChatGPT to catch obvious security holes, then fix them. Count real hours. Sixty focused hours is a useful planning figure. Don’t sell yourself short with “30 minutes” rhetoric.</p><h3>Final word</h3><p>Vibe coding changed the game. Two years ago I could not have built what I am building today without an ops team. Now I can ship real products in a fraction of that time, but only if I respect the gap. The shoreline is exciting. The crossing is where the work is. Learn to love both.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7e8b6d0b8da6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Extinction-Level Tech in the Hands of Cavemen]]></title>
            <link>https://medium.com/@thequickstartcreative/extinction-level-tech-in-the-hands-of-cavemen-e2c5841c085f?source=rss-13f43058462e------2</link>
            <guid isPermaLink="false">https://medium.com/p/e2c5841c085f</guid>
            <dc:creator><![CDATA[James A. Manning]]></dc:creator>
            <pubDate>Tue, 28 Oct 2025 11:04:16 GMT</pubDate>
            <atom:updated>2025-10-28T11:04:16.085Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9xQzZ0TMKhnJNiVt9TQd5g.png" /></figure><h3>We have the audacity to call this evolution.</h3><p>I follow a channel on Instagram called Mini Philosophy, and one of their posts got me thinking. The guy referenced biologist Edward Wilson and philosopher Toby Ord.</p><p>Wilson said humans have Paleolithic emotions, medieval institutions, and godlike technology. Ord took it further in The Precipice, saying that mismatch isn’t just theoretical—it’s existential. We’ve reached a point where we can wipe ourselves out, and the systems meant to govern us were never built for that kind of power.</p><p>And he’s right. America might be the clearest case study of how that plays out.</p><h3>The American Version</h3><p>Look at us. We’ve got the innovations: AI, global networks, biotech. But we’re still operating under laws written when there were thirteen states and human beings were property. The Constitution was built for farmers and merchants, not algorithms and nukes. Yet we act like democracy is self-sustaining, like it doesn’t require maintenance.</p><p>The deeper problem isn’t just outdated institutions. It’s what drives them: tribalism, scarcity thinking, fear of the other. White supremacy wasn’t a glitch; it was baked in. A system built on conquest over coexistence, power over equity.</p><p>And it’s not just America. Go anywhere (Africa, Asia, Europe, etc) and you’ll see the same pattern. Ethnic divides, religious wars, old hatreds repackaged with modern weapons. It’s human nature.</p><h3>It’s About Power</h3><p>When people talk about fear and tribalism, what they’re really talking about is power. Articles about “the browning of America” spark campaigns urging white families to have more children. Not because of love or culture, but because of panic. A fear that if they lose power, they’ll face what they once inflicted.</p><p>And if we’re honest, no group is immune. Hand any group unchecked control, and history tells you what happens. The problem isn’t a unique evil. It’s a shared wiring we refuse to confront.</p><h3>The Technology Problem</h3><p>That wiring now has access to tools that can end civilization. Nukes, AI, bioweapons. And information that spreads faster than truth.</p><p>Congress is still trying to write laws about AI like it’s 1962. They talk about “winning the AI race,” as if a few countries control it. But AI doesn’t respect borders. By the time they pass a bill, the technology they’re regulating will already be obsolete.</p><p>Or take healthcare. We have the wealth and the science, yet millions go bankrupt over insulin. Not because it’s unsolvable, but because the system is rigged for those who profit from dysfunction.</p><p>Race isn’t different. Gerrymandering and voter suppression are just cleaner versions of the old game: control who gets a voice.</p><h3>The Anonymity Problem</h3><p>Most people are decent face-to-face. Walk into a café and folks hold the door, say thank you, act like humans. But institutions give cover for cruelty. Bureaucracy is moral camouflage. You can vote for policies that harm people you’ll never meet and still feel righteous.</p><p>That’s the difference between a cop who knows his neighborhood and a faceless ICE agent in a mask. One is accountable. The other is a policy with a badge.</p><p>The further people are from the consequences of their actions, the easier it becomes to rationalize them.</p><h3>The Missed Intervention</h3><p>If there’s a fix, it starts with redesigning institutions to force human connection. To make it harder to dehumanize. To close the distance between decision and consequence.</p><p>But I don’t think we’ll take that path.</p><p>Obama said the arc of history bends toward justice. I think it bends toward entropy. Watch what happens this week when the government finds a way to pay the military but lets families on social programs starve. That isn’t dysfunction… it’s design.</p><p>This didn’t start with Trump. Congress has been rotting for decades. The system rewards those who thrive in anonymity and punishes those who tell the truth.</p><h3>The Long Decline</h3><p>I want to believe we’ll turn it around. I do. But the more I look, the more I think this ends with me off the grid, overlooking a winery in Oregon.</p><p>We’ll stumble forward until we find a “justified” reason to burn it all down. Because after Trump and Putin, there’ll be someone else. There’s always someone else. Another “other” to fear, another reason to fight.</p><p>Not because we’re evil, but because we’re unchanged. Still wired for small tribes, still organizing around power, still pretending technology will save us from ourselves.</p><p>We could fix it. We could build systems grounded in empathy and accountability. We could make equality more than a slogan.</p><p>But we won’t. The powerful like the game as it is, and the rest of us are too divided, too tired, or too distracted to force the change.</p><p>So I’ll keep doing what I can. Doing yeoman&#39;s work, fighting small battles, protecting the spaces that still feel human.</p><p>But the big picture? The godlike tools in Paleolithic hands, the medieval machinery running the show? It’s unfolding just as Toby Ord warned.</p><p>And I don’t think we’ll wake up in time.</p><p>The winery isn’t surrender. It’s the last form of resistance left where its just me and my wife drinking Pinot while whoever the hell is in charge at the time, burn this shit to the ground.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e2c5841c085f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Intelligence Trap: Why Most Companies Are Building the Wrong Kind of AI]]></title>
            <link>https://medium.com/@thequickstartcreative/the-intelligence-trap-why-most-companies-are-building-the-wrong-kind-of-ai-f7f0bb149b8e?source=rss-13f43058462e------2</link>
            <guid isPermaLink="false">https://medium.com/p/f7f0bb149b8e</guid>
            <dc:creator><![CDATA[James A. Manning]]></dc:creator>
            <pubDate>Mon, 20 Oct 2025 00:50:38 GMT</pubDate>
            <atom:updated>2025-10-20T00:50:38.895Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4UyUbPswT9jGmC3tPlW1VA.png" /><figcaption>Nano-banana image generation</figcaption></figure><p>Every company right now is asking: “How do we implement AI?”</p><p>That’s the wrong question. It’s like asking “How do we implement transportation?” when what you need to ask is: Should we walk, drive, or fly?</p><p>The real question is: <strong><em>How much intelligence does your process actually need?</em></strong></p><p>Most companies skip that step. They see a demo of a multi-agent system and assume sophistication equals value. They hire consultants who only know how to build one thing — complex orchestrated systems — whether the problem needs it or not. The result is the same every time: an overengineered project that costs too much, takes too long, and solves the wrong problem. I’ve seen companies spend six months and $150K building an “autonomous agent system” when a $5K workflow automation would have delivered results in three weeks.</p><h4>The Spectrum of Intelligence</h4><p>Intelligence exists on a spectrum. On one end, you have automation : systems that execute predefined rules. On the other, agents: systems that reason, adapt, and act with autonomy.</p><p><em>(This framework builds on ideas first outlined by Nate B. Jones, whose four-level model of AI autonomy inspired the structure here. My version treats those levels as a fluid spectrum rather than fixed stages.)</em></p><p>The mistake is thinking every problem sits at the far end of that spectrum. Most don’t. Most business processes don’t need AI to think; they just need it to act faster than a human can.</p><p>Here’s how to find where you really are on that spectrum:</p><p><strong>Level 1:</strong> Advisor / Co-Pilot — AI gives guidance, drafts, or suggestions, but the human executes. Think of ChatGPT helping you brainstorm or edit copy.</p><p><strong>Level 2:</strong> Tool-Augmented Automation — AI handles repetitive tasks through workflows or APIs. Humans set goals and review results. This is where most companies see their fastest ROI.</p><p><strong>Level 3: </strong>Human-in-the-Loop Intelligence — The system analyzes data and acts but pauses at critical decisions for human approval. Great for processes that involve judgment or risk.</p><p><strong>Level 4:</strong> Orchestrated Autonomy — Fully coordinated systems that plan, adapt, and execute with minimal oversight. Powerful, but only when your data, governance, and culture are mature enough to support it.</p><p>┌──────────────────────────────────────┐<br>│ Is the process repetitive and rule-based? │<br>└───────────────┬──────────────────────┘<br> │<br> YES ───┘──▶ LEVEL 2: AUTOMATION<br> │<br> NO ────▼<br>┌──────────────────────────────────────┐<br>│ Do you have structured data &amp; clear metrics? │<br>└───────────────┬──────────────────────┘<br> │<br> NO ────┘──▶ Fix data infrastructure first<br> │<br> YES ───▼<br>┌──────────────────────────────────────┐<br>│ Does the process require contextual │<br>│ judgment or handling exceptions? │<br>└───────────────┬──────────────────────┘<br> │<br> NO ────┘──▶ LEVEL 2: AUTOMATION<br> │<br> YES ───▼<br>┌──────────────────────────────────────┐<br>│ Can you afford mistakes during the │<br>│ learning phase? │<br>└───────────────┬──────────────────────┘<br> │<br> NO ────┘──▶ LEVEL 3: HUMAN-IN-THE-LOOP<br> │<br> YES ───▼<br>┌──────────────────────────────────────┐<br>│ Do you have governance infrastructure│<br>│ for oversight and accountability? │<br>└───────────────┬──────────────────────┘<br> │<br> NO ────┘──▶ LEVEL 3: HUMAN-IN-THE-LOOP<br> │<br> YES ───▼<br> LEVEL 4: ORCHESTRATED AUTONOMY</p><p>Before you add intelligence, prove the process needs it.</p><ul><li>If the steps are rule-based, start with Level 2.</li><li>If they require context or judgment, add a human-in-the-loop.</li><li>If your data isn’t structured or your metrics aren’t clear, fix that first.</li></ul><p>Because if you don’t have data, you can’t have intelligence.</p><h4>The Right Way to Scale</h4><p>Companies that win with AI all follow the same path. They start at Level 2 with simple automation.</p><p>They prove value fast. They build feedback loops so the system learns from outcomes. Then, when the data is strong and the workflows are stable, they graduate to Level 3.</p><p>The companies that fail skip these steps:</p><ul><li>They chase Level 4 autonomy before they’ve earned it.</li><li>Then they spend months rebuilding what they could have learned cheaply with smaller experiments.</li></ul><p>Technical debt piles up. Stakeholders lose patience. “AI strategy” becomes a cautionary tale instead of a competitive edge.</p><h4>Where to Start</h4><p>Ask one question:</p><p>“What’s one process where my team is doing repetitive cognitive work that could be augmented?”</p><p>Not eliminated… augmented. The best AI implementations don’t replace humans; they extend them.</p><p>Run that process through this test:</p><ol><li>Is it rule-based or judgment-based?</li></ol><p>2. Do we have measurable data?</p><p>3. Can we afford small mistakes while it learns?</p><p>Start with the simplest solution that works. Measure. Improve. Then move up the spectrum when the evidence says you’re ready.</p><p>That’s how you build capability that grows into intelligence, without falling into the trap of building intelligence before you need it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f7f0bb149b8e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Wrong Question About AI and Jobs]]></title>
            <link>https://medium.com/@thequickstartcreative/the-wrong-question-about-ai-and-jobs-ff6e274edc2f?source=rss-13f43058462e------2</link>
            <guid isPermaLink="false">https://medium.com/p/ff6e274edc2f</guid>
            <dc:creator><![CDATA[James A. Manning]]></dc:creator>
            <pubDate>Mon, 13 Oct 2025 00:47:45 GMT</pubDate>
            <atom:updated>2025-10-13T00:47:45.015Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Fh1y2od-37ca3F3OY0ELDA.png" /></figure><p>There&#39;s a Harvard study making the rounds that mapped 940 American occupations by two measures: how technically feasible it is to automate them, and how morally repugnant people find that automation. The researchers found that Americans support automating 58% of occupations when AI outperforms humans at lower cost, but a narrow 12% remains categorically off-limits due to moral repugnance, including caregiving, therapy, and spiritual leadership.</p><p>The researchers frame this as a distinction between &quot;performance-based concerns&quot; that fade as technology improves, and &quot;principle-based objections&quot; that persist regardless of capability. It&#39;s a tidy framework. It&#39;s also missing the point.</p><h4><strong>The Real Pattern</strong></h4><p>Look at the jobs people find morally repugnant to automate. Therapists. Teachers. Caregivers. Spiritual counselors. Now look at the jobs people are comfortable automating. Financial analysts. Search marketing strategists. Economists.</p><p>Here&#39;s what I&#39;m seeing. The difference isn&#39;t really about principles or how well the AI performs. It&#39;s about what we believe the human is actually doing in that job.</p><p>A financial analyst works on things. They analyze markets, crunch data, spot trends. Their job is to produce something: reports, predictions, recommendations. If an AI can produce the same output or better, then yeah, the AI can do the job.</p><p>A therapist works with people. They&#39;re not just producing a diagnosis or treatment plan. The whole point is that another human is sitting there with you, listening, present, trying to understand what you&#39;re going through. If you replace the human with AI, you might get the same technical output, but you&#39;ve lost the thing that actually mattered.</p><p>I&#39;m going to call this TO versus TOWARDS.</p><p>TO work: You&#39;re applying your expertise to something. Objects, data, systems, processes. The work is about producing a result.</p><p>TOWARDS work: You&#39;re applying your expertise toward someone. The work is about the human connection itself. Your presence as a human being is not just helpful. It&#39;s the entire point.</p><p>This is the distinction that matters. Not performance versus principle. Instrumental versus relational. TO versus TOWARDS.</p><h4>What the &quot;Moral Repugnance&quot; Is Really Detecting</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Zugg_HF3CZBBPrCA2_s5Sg.jpeg" /></figure><p>Here&#39;s what I think is happening when people say it&#39;s morally repugnant to automate certain jobs. They&#39;re not making a principled stand about which work must remain human. They&#39;re recognizing, maybe without being able to fully articulate it, that some work requires human presence not as a nice-to-have but as the entire mechanism by which the work functions.</p><p>Consider the nanny example. You could build an AI system that&#39;s perfectly safe, follows all developmental guidelines, never gets tired or frustrated. Perfect consistency, perfect patience. But when a parent chooses a human nanny despite these &quot;imperfections,&quot; they&#39;re not being irrational. They&#39;re recognizing that their child needs to learn how to be human from a human. The AI nanny&#39;s perfection isn&#39;t actually perfection. It&#39;s a category error.</p><p>The moral repugnance isn&#39;t squeamishness. It&#39;s sophisticated pattern recognition. We&#39;ve identified work where being human isn&#39;t a bug to optimize away but the feature that makes the work valuable.</p><h4>The Cultural Variable</h4><p>But here&#39;s where it gets more complicated. These boundaries aren&#39;t universal. They&#39;re culturally constructed and economically contingent.</p><p>Ask Americans about automating school psychologists and you&#39;ll get high moral repugnance. Ask parents in a country with a massive shortage of mental health services, and that calculation might shift. The moral objection persists until economic reality makes it untenable. Then we retrofit our moral frameworks to accommodate what&#39;s already happening.</p><p>Different cultures will draw these TO versus TOWARDS boundaries differently. Some might see teaching as primarily instrumental (transmit information efficiently). Others see it as fundamentally relational (form young minds through connection). Economic necessity will push cultures to accept more TO framing when they can&#39;t afford the TOWARDS version.</p><p>Who you ask matters too. Survey construction workers about their work and many would argue their craft should score high on moral repugnance. Not because of principles, but because they understand that what they do involves judgment, coordination, problem-solving, and pride in the finished work. That&#39;s TOWARDS work, even if it&#39;s not towards a specific person. It&#39;s towards their crew, their craft, their sense of mastery.</p><h4>Most Jobs Are Both</h4><p>This is the part the Harvard research doesn&#39;t capture. Most jobs aren&#39;t purely instrumental or purely relational. They&#39;re a mix.</p><p>A doctor has TO work: reviewing test results, updating records, diagnosing standard conditions. This is work AI can potentially do better. But a doctor also has TOWARDS work: delivering difficult news, understanding the patient&#39;s life context, making judgment calls when the data is ambiguous. This is work that requires human presence.</p><p>A teacher has TO work: grading, lesson planning, administrative tasks. And TOWARDS work: responding to a student&#39;s confusion in real-time, noticing when someone is struggling emotionally, building confidence through relationship.</p><p>The research plots entire occupations as either automatable or not. But that&#39;s not how work actually functions. Every role contains both instrumental tasks and relational tasks. The question isn&#39;t whether the job gets automated. The question is which parts get automated and which parts become more valuable because they&#39;re distinctly human.</p><h4>The Better Question</h4><p>So instead of asking &quot;Will AI take my job?&quot; we should be asking &quot;Which parts of my work are TO and which parts are TOWARDS?&quot;</p><p>The TO parts are fair game for automation. In fact, automating them might be good. It frees up time and energy for the work that actually requires human qualities: judgment, presence, connection, creativity.</p><p>The TOWARDS parts are where we should be doubling down. Bringing our full selves to the work, because our full selves are the point.</p><p>This reframes the automation anxiety. It&#39;s not about protecting jobs from AI. It&#39;s about understanding which aspects of work are uniquely human and structuring our roles to emphasize those aspects.</p><h4>What This Actually Means</h4><p>The researchers found that support for automation nearly doubles when AI is described as outperforming humans at lower cost. That&#39;s not principles eroding. That&#39;s basic economics.</p><p>But that stubborn 12% that stays off-limits? That&#39;s not arbitrary. That&#39;s recognition that some work is fundamentally relational. You can&#39;t optimize it the same way you optimize instrumental work. Making therapy more &quot;efficient&quot; might actually make it worse. Scaling caregiving by removing the human relationship might solve a capacity problem while creating a care problem.</p><p>The task ahead isn&#39;t figuring out which jobs to protect from automation. It&#39;s figuring out how to separate the TO from the TOWARDS in every job, automate the former, and amplify the latter.</p><p>Because here&#39;s what the research really shows: we&#39;re not afraid of AI doing work. We&#39;re afraid of AI replacing human presence in contexts where human presence is the entire point.</p><p>And that fear isn&#39;t irrational. It&#39;s the most rational thing in the study.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ff6e274edc2f" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>