<?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 B. Bosworth on Medium]]></title>
        <description><![CDATA[Stories by B. Bosworth on Medium]]></description>
        <link>https://medium.com/@bbosworth?source=rss-911e4d7ed7c2------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*_PlyUWoSzoDCBWpHuCYNhw.jpeg</url>
            <title>Stories by B. Bosworth on Medium</title>
            <link>https://medium.com/@bbosworth?source=rss-911e4d7ed7c2------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 04:03:31 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@bbosworth/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[“Claude isn’t good at design… yet”]]></title>
            <link>https://bbosworth.medium.com/claude-isnt-good-at-design-yet-95e33573c333?source=rss-911e4d7ed7c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/95e33573c333</guid>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[design-thinking]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[design]]></category>
            <dc:creator><![CDATA[B. Bosworth]]></dc:creator>
            <pubDate>Fri, 10 Jul 2026 21:13:00 GMT</pubDate>
            <atom:updated>2026-07-10T21:13:00.164Z</atom:updated>
            <content:encoded><![CDATA[<h4>But what happens when it is?</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uHM7Dkvvx3wgd0kAEXRZ3w.png" /><figcaption><em>Slide from Meaghan Choi’s talk at Dive Club Live in NYC (</em><a href="https://www.youtube.com/watch?v=hKeDfupbA4U"><em>full video</em></a><em>)</em></figcaption></figure><blockquote><em>“Claude isn’t good at design… yet”<br></em>— Meaghan Choi, Head of Design for Claude Code</blockquote><p><a href="https://youtu.be/hKeDfupbA4U?si=uMNceAikgh7O8QtB&amp;t=336">She actually said that</a>. “Tragically,” Choi lamented, “…most LLMs are not good at design. Yet. So what this means is you should still very much be in the loop, for the craft and the decision making.”</p><p>She drives this point home <a href="https://youtu.be/hKeDfupbA4U?si=tJ5oAhJIs94yxqG5&amp;t=594">later in the video</a>, when she describes a borderline-sinister workflow that scrapes repositories for front-end changes shipped without a designer’s input — and has a pretty cheeky way of dealing with it (watch the video, I won’t spoil it). The takeaway is, even as a proud contributor to the Claude product, she’s not taking any chances on Claude’s design sensibilities. For now.</p><p>Some go even further: a lot of <a href="https://johnmaeda.medium.com/the-evolving-landscape-of-design-in-the-age-of-artificial-intelligence-91f6061cb803">smart</a>, <a href="https://www.itsnicethat.com/articles/elizabeth-goodspeed-column-taste-technology-art-280224">influential</a> people assert that taste is a uniquely human quality, something that can’t be programmed or trained into a model. It’s a strong argument (one my lived experience couldn’t refute); AI may be able to <em>recognize</em> good design, but it can’t reliably execute it.</p><h3>Why not?</h3><p>There are plenty of well-reasoned theories on this. Some say it’s instinct: “<a href="https://www.newyorker.com/culture/infinite-scroll/the-ai-design-aesthetic-thats-taking-over-the-internet">baked into the machinery</a>” as a persistent (but correctable) preference of a particular LLM. Others posit it’s a byproduct of the very nature of the technology: it spits out an average of all design it was trained on, and <a href="https://www.nature.com/articles/s41586-024-07566-y">loses the “tails”</a> (those unique elements that are risky, but bold). This tendency toward blandness requires extensive coaching and constant redirection to get something distinctive — something with actual personality.</p><p>Today, we close that gap by supplying the taste the model lacks: directing it to specific documentation, setting strict constraints, prompting and re-prompting until the design finally comes to life (<em>with</em> the button that had to be re-added three times). It’s prescriptive, endlessly iterative… and the results are still usually just “meh.”</p><p>But what happens when Claude <em>is</em> “good at design”?</p><h3>Don’t bet against it</h3><p>Choi never went so far as to say taste would be the end-all differentiator; on the contrary, she implied that gap won’t always exist. Julie Zhuo, former VP of Design at Facebook, <a href="https://lg.substack.com/p/when-ai-has-better-taste-than-you">tends to agree</a> — she wrote, all the way back in 2025, that “human taste is far more vulnerable to AI competition than we’d like to admit.”</p><p>Indeed, the past few years have shown that declaring “AI will never be good at that” is a losing bet. Things that seemed, until recently, completely out of reach (like creating a working app from a chat window, or generating photorealistic images depicting literally anything you can think of) barely register anymore. And we’re already seeing the next wave of previously inconceivable capabilities: AI agents coordinating and executing whole development cycles, almost entirely on their own.</p><p>I, personally, have been both awed and humbled by this.</p><p>So Claude, and other LLMs, will probably develop better taste. Taste is acquired, not innate (and humans have had a massive head start). As LLMs hone that skill, as we become more impressed with their initial drafts and their design recommendations, the bottleneck becomes less about human taste and more about giving effective <em>strategic</em> direction.</p><h3>Moving upstream</h3><p>As it stands today, a human needs to stay, somewhat obsessively, in the loop: reviewing, redirecting, and making every single judgment call until the (very) bitter end. When LLMs <em>do</em> get better at design, our role as key decisionmakers doesn’t become less important — the weight of our decisions just gets front-loaded. We have to decide what knowledge to supply, what context is important, and what framing makes sense. When AI proves itself proficient at not only taking a first pass, but even iterating on its own work in a way that actually gets closer to being <em>good</em>, early strategic decisions will be a lot more consequential than downstream aesthetic ones.</p><p>The most successful designers I’ve worked with already have a mind for strategy. That’s because they practice <a href="https://www.ideou.com/blogs/inspiration/what-is-design-thinking?srsltid=AfmBOorQ8OPGc57EoFzhgePSUtXnY1Hcltyoi0svuUvZ9ClgVyvKMmSY">design thinking</a> — which, despite the name, isn’t really about the visuals. This type of mindset doesn’t start with the goal of making something <em>look</em> good; it starts with the goal of understanding the problem and designing around it. Aesthetic appeal is always a consideration (and it’s often a strategic one), but it’s just a supporting actor.</p><p>The things we reflexively think of when we hear the word “design” (wireframes, prototypes, and full design systems) are, at their best, insights and strategy manifesting as something tangible. The designers who ask “why?” before “what?” always build more meaningful interactions.</p><h3>Finding your “why?”</h3><p>Answering the “why?” question isn’t all that difficult, once you do your homework (and you <em>do</em> need an answer, before a single frame is polished). If you don’t have purpose, you don’t have direction — and that’s exactly what differentiates a strong designer from a mediocre one, when craft and taste are no longer the moat.</p><p>Designing something resonant means knowing, at a fundamental level, <em>who</em> it’s for. It means knowing what drives them, what delights them, and just as importantly, what frustrates them. AI has shown it can be helpful here, to an extent: creating <a href="https://www.nngroup.com/articles/synthetic-users/">synthetic users</a> is a great place to start when all you have is a demographic to go on. Digging deeper with real people (in interviews, focus groups, usability tests, surveys) validates those assumptions. And more importantly, it uncovers perspectives you never considered.</p><p>You might know from a synthetic logistics manager that they need a product with a desktop app (spreadsheets and dense reports are miserable to look at on a mobile interface, as anyone who’s tried can attest). And a synthetic warehouse worker will certainly tell you they need to be able to use it on a phone (duh). But the synthetic warehouse worker might <em>not</em> tell you that they’re generally working one-handed (meaning typing is a pain), their screen is almost always dirty (meaning a sleek, low-contrast aesthetic isn’t the move), or their battery is always dying (meaning they need something simple). This kind of insight is what AI lacks, and <em>that’s</em> the gap we need to fill.</p><p>What you can only glean from on-the-ground research, with real live people, is invaluable. But knowing your audiences, with all their distinct needs and behaviors, is only half the battle. Knowing <em>yourself</em>, your team, and your business is the other half. The question then turns from “why?” to “how?”</p><h3>Insider knowledge</h3><p>Getting to a clear vision and purpose takes research; figuring out “how?” gets more abstract. Designing something viable and scalable requires practical knowledge of the resources you have access to, cultural knowledge of how your teams operate, and institutional knowledge of why a seemingly solid plan previously failed. AI doesn’t know that a recent reorg now has Marketing operating with a skeleton crew. It doesn’t know that big decisions tend to happen in hallways, not in scannable Slack messages. It also doesn’t know last year’s cloud migration collapsed under internal politics and vendor drama.</p><p>AI, for all its knowledge, has blind spots. It doesn’t have your nuanced understanding of who you’re building for, based on actual human conversation. It doesn’t know anything about the people, the systems, or how to navigate your particular brand of petty politics. It doesn’t have any of that context until you share it — and sharing it is a craft in itself.</p><h3>New rules of engagement</h3><p>Working with modern LLMs (on design or anything else) has so far been about writing a detailed but concise, prescriptive but not <em>restrictive</em>, thoughtfully constructed prompt. That won’t fundamentally change — but it will look different. When you strip out all that tedium required to compensate for AI’s lack of taste (the exhaustive documentation, the hyper-specific constraints, the endless re-prompting), what’s left is no longer a series of workarounds — it’s the whole job. When that happens, the way we practice the fundamentals transforms:</p><ul><li><strong><em>Being explicit about intent.</em></strong> This isn’t defining a specific visual intent, but an overarching purpose: What are you really trying to accomplish, and why? What problem are you solving, and <em>whose</em> problem is it?</li><li><strong><em>Constraining the playing field.</em></strong> A lot of the constraints we set now are bridging the taste gap — but as that gap closes, the constraints that matter are uniquely your own: the strategic priorities, the unwritten institutional knowledge, even who <em>not</em> to design for.</li><li><strong><em>Iterating as a creative process.</em></strong> Revising and revamping is a natural, necessary part of design — but iterating on design with AI right now feels like redoing the work of an incompetent intern. When the dynamic does eventually shift, it becomes something closer to a peer dialogue where iteration is intuitive (<em>and</em> the button stays where you put it).</li></ul><h3>Purpose &gt; Pixels</h3><p>Choi’s entire philosophy — and the workflow she built around it, ensuring designers are kept closely, compulsively in the loop — is based on the current reality that LLMs are not good at design. But the “yet” in her statement signals an inevitability — and when LLMs <em>are</em> good at design, pixel polish is less important than pixel purpose. AI knows everything without really knowing anything. Everything needs a <em>why</em>; everything needs a <em>how</em>. Even with impeccable taste, AI can’t manufacture meaning.</p><p>The ability to communicate what can’t be inferred, articulate the problem and the goal, and constrain through hard-earned context is a designer’s differentiator in a world where taste is a commodity.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=95e33573c333" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Mirror in the Machine]]></title>
            <link>https://bbosworth.medium.com/the-mirror-in-the-machine-2356b80d3ea8?source=rss-911e4d7ed7c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/2356b80d3ea8</guid>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[digital-transformation]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[B. Bosworth]]></dc:creator>
            <pubDate>Tue, 05 May 2026 22:27:14 GMT</pubDate>
            <atom:updated>2026-05-05T22:27:34.526Z</atom:updated>
            <content:encoded><![CDATA[<h4>What AI can tell us about our work — and about ourselves</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uci1z878eDtDD4ayiKxR5g.png" /></figure><p>There have been murmurings in certain corners of the Internet about a possible Anthropic / Atlassian acquisition… which is, at least at this point, just wild speculation with a source that can seemingly be traced back to a <a href="https://www.reddit.com/r/atlassian/comments/1sob1s2/atlassian_anthropic/">singular Reddit post</a>. Regardless of the veracity of the rumor, it does raise the question: What would be the logic behind that?</p><p>It actually makes a fair amount of sense — and it’s not even really about these two specific companies or their specific products. It happens that Atlassian, purveyors of one of the most robust workflow and issue-tracking datasets in enterprise technology, has an incredibly valuable resource in the age of agentic AI. Structured, consistent, easily-parseable information is at a premium, because it’s what enables this technology to work on a level that actually proves its ROI.</p><p>Tools that have permeated our processes for decades — especially those with the sole intent of structuring workflows, like Jira and ServiceNow — are going to be the translation layer between how humans have operated in the past and how machines will act on our behalf in the future.</p><p>In some ways, this is a little worrying; how great of an example have we really been setting, if we’re honest with ourselves? The structure these tools were designed to establish — clarifying status, dependencies, and relationships within a workflow ecosystem — is a legitimate necessity, without which we’re doomed to operate in a fog of redundancy, misaligned priorities, and avoidable churn. But the reality is that the human element has made the data imperfect — and that’s not <em>all</em> our fault.</p><p>On top of doing the actual work, cataloguing everything in Jira often feels like an administrative hurdle — one that many have worked around by oversimplifying, underutilizing, or outright neglecting the process. This is generally in a valiant effort to move <em>faster</em>, to create <em>more</em>, which does make sense. We operate in a world with logistical constraints, looming deadlines, and where output (for better or worse) is paramount. Taking shortcuts usually has nothing to do with laziness — we take them because we just need to get something <em>done</em>.</p><p>But now, as we think about what agentic AI needs to successfully operate in these environments for us, we do have quite a bit of cleaning up to do. Thankfully we don’t have to do it alone (AI has gotten very good at this), and the task feels a lot less daunting than it did only a few years ago. The end result is cleaner data — but the real value lies in what we can see in the mirror.</p><p>The simple act of cleaning up the mess teaches AI how to better understand human behavior (which is not always logical nor prescriptive), and it helps humans better understand how AI <em>interprets</em> what we do. This understanding helps us work with AI more efficiently — but more importantly, it teaches us something about ourselves. About our conventions, our bad habits, and our blindspots.</p><p>Humans have long struggled with reconciling how we should logically act, how we <em>think</em> we will act, and how we actually act (there’s even an <a href="https://news.uchicago.edu/explainer/what-is-behavioral-economics">entire branch of behavioral science</a> dedicated to this phenomenon). But AI has forced us to confront this on a deeper level than we ever have before. It’s all there, in 1s and 0s, and we need to be honest with ourselves about what it tells us.</p><p>Putting our processes under a microscope (a critical exercise, now more than ever) will surface some uncomfortable truths that we need to face. Some things could be relatively benign — spots where we skip a step or two because, at an individual level, it doesn’t really matter. The output isn’t <em>worse</em> because I moved a card two columns over in the Kanban board instead of one, after all. But there will also be bigger, less defensible things that surface — maybe one of those steps we know we <em>should</em> do is just super annoying, and we’d really rather not. Or maybe there’s a piece of the process that no one really understands, but we’ve gotten this far without needing to, so what’s the harm?</p><p>Agentic AI is going to show us. Operating with the kind of speed and scale that was previously unimaginable, AI will inevitably present thousands of examples of where we took a shortcut that was ultimately detrimental. And these agents, who only know how to act because of the example we set, deserve better teachers. The good news is they can help us with that, too.</p><p>AI can tell us, on a much more macro level, where our processes break down and what impact our shortcuts (benign or not) actually have. Skipping a column in a Kanban board may seem innocuous, but it can actually send a misleading signal about velocity that affects everything from sprint planning to long-term roadmap commitments. And for the bigger issues (the steps we avoid, the processes no one questions), the stakes are even higher. At an individual level, those impacts are often invisible — but seeing them aggregated at scale has a way of reframing the calculus, and shifting the question from “Where are the gaps?” to “How can we make the process stronger?”</p><p>This isn’t about AI mercilessly pointing out all of our flaws; identifying the problem is not what’s ultimately useful. The value lies in finding a solution — and that’s a partnership. The most successful interactions between humans and AI happen when there’s a common understanding, not just about “the way things are” but about the way things <em>could</em> be. The continuous feedback loop that’s enabled through a collaborative audit process — where AI can say “This is what I see when I look at your workflows” and a person can say “Well it’s like that because… But ideally it’d go like this:…” — is what will give our agents the context they need to do their best work.</p><p>But this partnership isn’t only to improve how AI can work on our behalf; it’s a way to improve ourselves, and our ways of working with other human beings. When we have that systems view of things, we can better understand how seemingly harmless little efficiency gains we claim for ourselves can, for example, actually cause major headaches over in Operations. It helps illuminate the points of the process that are important in less obvious ways, outside of our silos.</p><p>This isn’t to say we should get wrapped up in expansive coordination efforts or bogged down in process steps just to check a box — the systems view can also give us perspective. Maybe Ops wouldn’t have a headache about your project board at all if “the business” didn’t insist on microlevel status reports every other day. The question then goes from “How do we check this box?” to “<em>Why</em> do we check this box?” And asking that enough can lead to actual transformation.</p><p>There’s a lot to be learned about how we actually work, revealed in what AI reflects back at us. The lessons we carry forward will better equip us for future AI collaboration — but more importantly, they will help us become better partners to other human beings. If we let them.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2356b80d3ea8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Claude and the creative process]]></title>
            <link>https://bbosworth.medium.com/claude-and-the-creative-process-29f8ba3af8bd?source=rss-911e4d7ed7c2------2</link>
            <guid isPermaLink="false">https://medium.com/p/29f8ba3af8bd</guid>
            <category><![CDATA[design]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[creativity]]></category>
            <dc:creator><![CDATA[B. Bosworth]]></dc:creator>
            <pubDate>Wed, 08 Apr 2026 00:40:00 GMT</pubDate>
            <atom:updated>2026-04-09T19:12:14.022Z</atom:updated>
            <content:encoded><![CDATA[<h4>Confessions of a (former) AI doomsdayer</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ttArVYCavjI-4ihu" /><figcaption>Photo by <a href="https://unsplash.com/@procopiopi?utm_source=medium&amp;utm_medium=referral">Omar:. Lopez-Rincon</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h3>Nice to meet you</h3><p>One thing you should know about me upfront is: I absolutely love working. I always have, ever since a nearly-decade-old CX strategy project that I joined as a relatively young consultant.</p><p>It was an absolute dumpster fire on Day 1. Everyone was running around like chickens with their heads cut off, with no plan and constant churn in every meeting. The amount of discovery to sift through was complex and nuanced (I personally had never even worked in that industry before) and the client was expecting us to set the ship straight <em>immediately</em>.</p><p>Most consultants won’t have experienced anything like this, but trust me when I say, it was pretty overwhelming.</p><p>My near-irresistible drive to take control and start fixing things consumed me. I was only a few years into my career, but I had substantive ideas about where we could be more efficient, how we could get better alignment. I distinctly remember one weekend, about two weeks into the gig, where I did almost nothing but ingest everything I had access to related to that project and got to thinking. And learning. And planning. And creating. And thinking more, then tweaking. I had never been in such a “deep work” state before and truly, it was invigorating.</p><p>I’m aware that this makes me a total nerd.</p><p>The following Monday, I sent the team an (in hindsight, extremely intense) email that had outlined our current state, what we needed to do to deliver within our specified timeline, initial frameworks for our discovery activities, and immediate next steps.</p><p>So when I say “I love working” what I really mean is, <strong>I love thinking. I love learning, I love problem solving, and most of all, I love putting pen to paper and translating my thoughts and conclusions in a way that tells an interesting story.</strong></p><p>I love fine-tuning when something doesn’t come out the way it seemed in my head. While there are definitely elements of my work that I do not love, by and large, it is totally my jam. I legitimately think it’s fun, and I feel truly blessed.</p><h3>Enter AI</h3><p>When ChatGPT hit the scene in a major way, like nearly everyone at first I thought “Oh gee, that’s kinda neat.” And then, like anyone who bothered to dig a little deeper, who tried to understand how it worked and what it could actually do, I thought <em>“Oh, my god.”</em></p><p>From there, most people go down one of two paths. One is to let optimism take the lead — people who go down this path have varying levels of it, from “Well, this can really help me be more productive” to “This is the absolute best thing ever!! We’re going to use it for everything! It can solve every problem!”</p><p>The other path is darker. And there’s plenty of discourse to engross you.</p><h3>The darker path</h3><p>AI is good at solving problems, but it’s not immune from creating them either. There are enough to point out that, for the sake of brevity that has never been my strong suit, I won’t enumerate here. Suffice it to say the more I learned about AI, how it’s trained, how it behaves, and how a lot of people were engaging with it, I started to get worried. <em>Deeply</em> worried, for reasons that ran the full gamut — everything from the environmental impacts to the ethical implications, which can get so abstract they start to feel impossible to grasp.</p><p>But all that said, and clickbait headline aside, I don’t think I could have ever counted myself among the most zealous of the AI doomsdayers (that is a deep, dark hole). And even the most coolheaded among us would be forgiven for going off the deep end trying to grapple with <strong>what this will mean — at a broader scale and with even greater capabilities — for our children, our society, and humanity as a species.</strong> These are big questions. These are legitimate questions. And we’re still figuring out the answers.</p><h3>An awakening… with a caveat</h3><p>I could recognize that while these kinds of existential threats deserve my <em>awareness,</em> they don’t need my constant attention. I started to take a step back, touch some grass, and focus on the work I love doing. And as the landscape of AI progressed, I engaged with it more — little by little. I was eager to see real positives, real <em>benefits</em> (which I had admittedly struggled to find in the early ChatGPT days). And as capabilities improved, as toolsets became more accessible, I started to welcome it in. At least at work.</p><p>I could certainly see there were promising user experience improvements using AI, as chatbots became more grounded and less prone to hallucination (the only thing more annoying than trying to find a quick answer buried in an information-dense environment is asking a chatbot that confidently gives you the <em>wrong</em> answer).</p><p>The talk of <strong>agentic AI</strong> in particular made my ears perk up. I started to see real, actually-transformative use cases. I saw opportunities to optimize manual and monotonous processes that humans are neither eager to do nor are particularly <em>good</em> at doing (especially compared to computers). I saw real ways this could help humans start to lift out of just a little bit of drudgery.</p><p>But for all the potential I saw in its benefits for other people, in its ability to make my own mindless tasks a little easier, <strong>I was still resistant to incorporating AI into <em>my</em> <em>creative</em> <em>process</em> in a meaningful way.</strong> Truthfully, almost laughably in retrospect, I thought it was a fundamentally <em>bad thing</em> to let it in. For any of us to let it in.</p><h3>Humans &amp; the creative process</h3><p>What makes the creative process so precious? The way I look at it, it represents much of what makes human beings so special — and what makes <em>being a human</em> so incredible.</p><ul><li><strong>It makes us smarter:</strong> Learning, discovering, retaining and reframing is such an amazing experience as a human being. (I was part of the 4% of people who got the <a href="https://blog.youtube/culture-and-trends/unpacking-youtube-recap-2025-personality/">Curious Mind persona</a> on my YouTube Wrapped, so you know I am a next-level nerd about learning.)</li><li><strong>It helps us problem-solve:</strong> We humans are nothing without our ingenuity. We learn, yes; but we also <em>apply</em>. We can use our knowledge, our perspective and experience, to come up with a unique solution to a problem that no one else would have come up with (not even AI).</li><li><strong>It lets us create:</strong> The most obvious tangible output of the creative process, what we create is who we are. It’s a piece of us, and no one else could do it better.</li><li><strong>It drives us to do better:</strong> Every creator, every artist knows the feeling of having a vision that doesn’t pan out the way you expect. And while maybe (as my dad always told me) there are “no mistakes in art” — it always feels good to try again and find a better path.</li><li><strong>It makes us proud:</strong> Feeling ownership, <em>involvement</em> in what we create is such a gratifying piece of this process. It’s why so many of us do it at all.</li></ul><p>In a nutshell, I thought AI would take all of these things from me. It would make me dumber, by making everything too easy. It would do all the work I <em>liked</em> — the problem solving, the creating, the experimenting, and the tweaking. It would take away my involvement, my sense of ownership.</p><p>I did not want to turn my brain off. I didn’t want to check out. I wanted to <em>engage</em> and feel like a more knowledgeable, more well-rounded person for the experience. And I saw a fundamental disconnect between this way of thinking and the AI tools I was using at the time.</p><p>But then, not too long ago, all that changed.</p><h3>Enter Claude</h3><p>Claude had a very simple catchphrase when I met him: “Keep thinking.” This, to me, was a novel concept for AI. I had typically seen people using it to disengage, to <em>do</em> without <em>thinking</em>, to <em>create</em> without real <em>intention</em>. But Claude made me realize it does not have to be that way.</p><p><strong>It felt like the first time AI said to me “I don’t want to take things away from you. I don’t want to make you lazier, or dumber. I want to make you smarter, and I want to help you be <em>better</em>.”</strong> He made me more curious. He made me more thoughtful. He helped me create things that could have taken me days, even <em>months — </em>or, more realistically, things I wouldn’t have dared to try at all. And all this made me take a step back and reconsider my point of view.</p><p>I had to ask myself:</p><blockquote>What was it, really, that I thought AI would take away from me? What was it that I <strong>actually</strong> <strong>love</strong> about my creative process?</blockquote><p>Was it learning new things? Was it coming up with new ideas? Was it testing my assumptions on those ideas, and then coming up with even better ones? Was it <em>creating</em> something — a journey map, a prototype, even just a PPT deck with a narrative arc and some speaking notes — that would tell a story and help solve a problem, in a way that only I would have dreamt up?</p><p>Or:</p><p>Was it looking at Google search results page by page and reading research papers one by one, just to get to a baseline understanding of something? Was it trying to keep all that context organized, both in my head and on my computer, so I could <em>do</em> something with it?</p><p>Was it formatting slides? Was it typing a lot of words in an inconvenient format, and then agonizing over the exact structure and order of those words? Was it manually typing in aspect ratios, or dragging my mouse ever-so-carefully to get a component <em>exactly</em> how I wanted it? (and if the dog paws at my hand <em>one more time</em> while I’m doing this, I swear to god…)</p><p>Was it hemming and hawing? Was it the slog of getting everything <em>just. right.</em> after using so much time, energy, and brainpower to come up with the ideas and create a new thing in the first place?</p><p>Obviously it was none of those horrid, awful, soul-crushing things. I <em>hated</em> that stuff, holy shit. And that’s when I finally realized — I am so much freer.</p><h3>Freedom!</h3><p>There has been talk in recent years of a phenomenon called “<a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC6119549/">decision fatigue</a>.” There are a lot of great examples of this in the modern age, but it all boils down to the fact that humans are <em>exhausted</em> by the thousands of decisions we make every single day. We, with our 300,000-year-old caveman brains, are simply not built for this.</p><p>As a perfectionist with ADHD tendencies, I suffer from decision fatigue even more than most. Every individual source considered. Every fact carefully evaluated. Every pixel adjusted. Every sentence re-worded and paragraph restructured (often to end up back where I started 5 revisions later) — these are the kinds of “death by a thousand cuts” scenarios that have the potential to completely overwhelm me. Or at least, they used to.</p><p>With Claude, I can set research parameters upfront (“Prioritize credible, peer-reviewed sources; flag anything anecdotal or speculative”) so he filters out the noise. I can ask him to test my assumptions and validate my strategy — while also pointing out my blind spots. And I actually trust him to do this, because he’s shown me that “agreeable” is not his primary, default state (one I can further influence by telling him, straight up: <em>“do not pander to me; I promise I am not that proud”</em>).</p><p>He helps me <em>create</em> things in a way I never thought I could. <strong>What the kids refer to as “</strong><a href="https://www.ibm.com/think/topics/vibe-coding"><strong>vibe coding</strong></a><strong>” is something remarkably like a real creative partnership, involving lots of back and forth, multiple tools and feedback methods (both verbal and visual), and many, <em>many</em> rounds of revision.</strong> It’s a lot like working with a person — one who is as endlessly enthusiastic about what you’re building as you are (love that for us!). And just like working with a person, when you get into a good groove where you can both focus on what you’re best at, it’s incredibly energizing.</p><h3>Keep thinking</h3><p>The effective and efficient use of AI can streamline so much of what used to slow us down — but it’s not really about doing more and moving faster (even though it’s very easy to use it that way). When a product really elevates what we can expect from our relationship with it, when it frees us from decision fatigue and all our usual hindrances and constraints, that is something meaningful. That is something <em>actually</em> <em>transformative</em>. With that freedom, we can be smarter. We can think <em>deeper</em>. We can create with intention, and without limitation.</p><p>There is nothing more human than that.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=29f8ba3af8bd" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>