<?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 Jelmar Groot on Medium]]></title>
        <description><![CDATA[Stories by Jelmar Groot on Medium]]></description>
        <link>https://medium.com/@jelmar_10xbetter.ai?source=rss-db308144e40d------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*xt_mXyhVKBtjy_VtNmMB1Q.jpeg</url>
            <title>Stories by Jelmar Groot on Medium</title>
            <link>https://medium.com/@jelmar_10xbetter.ai?source=rss-db308144e40d------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 09:06:32 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@jelmar_10xbetter.ai/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[What AI can do — and what it can’t]]></title>
            <link>https://medium.com/@jelmar_10xbetter.ai/what-ai-can-do-and-what-it-cant-8b26b751f107?source=rss-db308144e40d------2</link>
            <guid isPermaLink="false">https://medium.com/p/8b26b751f107</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[agi]]></category>
            <category><![CDATA[coding]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[writing]]></category>
            <dc:creator><![CDATA[Jelmar Groot]]></dc:creator>
            <pubDate>Mon, 30 Dec 2024 08:44:40 GMT</pubDate>
            <atom:updated>2024-12-30T08:44:40.074Z</atom:updated>
            <content:encoded><![CDATA[<h3>What AI can do — and what it can’t</h3><figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*HKkMHm5FbX63rEIcBWwKbg.png" /></figure><blockquote>“The Electric Monk was a labour-saving device, like a dishwasher or a video recorder. Dishwashers washed tedious dishes for you, thus saving you the bother of washing them yourself, video recorders watched tedious television for you, thus saving you the bother of looking at it yourself; Electric Monks believed things for you, thus saving you what was becoming an increasingly onerous task, that of believing all the things the world expected you to believe.”</blockquote><p><em>From: Dirk Gently’s Holistic Detective Agency, by Douglas Adams</em></p><p>I start this newsletter with this timeless quote because it highlights something fascinating about our relationship with technology. The idea that something as uniquely human as believing could be outsourced to a machine seems absurd — yet here we are, half-expecting AI to think, create, and understand for us.</p><p>Today’s AI tools are, in a way, our Electric Monks. Your email client suggests it can capture your tone and intent. Your coding assistant claims it can architect your solutions. Your writing tool offers to generate your reports. Each promise pushes us to question what we can truly delegate to machines.</p><p>As a writer, this fascinates me: why can AI now code at the level of top programmers, yet fail to produce meaningful literature? The answer lies in patterns.</p><h4>The Pattern Paradox</h4><p>OpenAI recently announced their O3 model, claiming coding that would place it among the world’s top 200 programmers. While we await real-world demonstration of these, the contrast is telling: the same AI that could potentially match elite programmers still struggles to understand why a joke is funny.</p><p>This paradox explains why people tend to both overestimate and underestimate AI’s abilities. The key to understanding AI is not about how it measures up to human intelligence, but how well it handles patterns. Anything that can be reduced to patterns, no matter how complex, is potentially within AI’s grasp. (Though sadly, despite my wishful thinking, I haven’t found an AI that could handle my family patterns over Christmas.)</p><p>As a former yoga teacher, I’ve spent years helping people break free from limiting patterns — emotional reactions, thought loops, habitual behaviors. Yet here’s AI, excelling precisely because it masters patterns rather than breaking them.</p><p>This isn’t a coincidence.</p><p>While patterns are incredibly valuable — they help us learn, grow, and function efficiently — many of humanity’s most precious achievements come from transcending them. Our greatest breakthroughs, our deepest connections, our most meaningful experiences often emerge when we step beyond established structures.</p><h4>Patterns at Work</h4><p>Your workplace is full of patterns. The data analysis you run each month. The code you write to automate tasks. The research you compile for presentations. AI handles these beautifully because they follow recognisable structures.</p><p>Googles new deep research model (Gemini 1.5 Deep Research) demonstrated this so effectively to me that I renewed my subscription, despite already using Claude for most of my AI needs.</p><p>When writing articles requiring extensive research, the new model replaced my need of opening twenty browser tabs and spending hours collating disparate pieces of information. It created grounded, factual foundations for original writing. No hallucinations. No made-up sources. Just solid research synthesis.</p><p>For a copywriter this is awesome. It’s way beyond having a sketchpad full of ideas (although I still need that too); it’s a deep background document on my chosen topic that I can have in my hands in a matter of minutes.</p><p>So of course it works great for this. Analysing vast amounts of information is probably the one of AI’s most important use cases at the moment.</p><p>But then there’s everything else.</p><p>The email where tone really matters. The presentation that needs to inspire. The conversation that builds trust between teams. All of these things rely on patterns, but only become truly meaningful if they also somehow transcend the structural elements that enable them to be constituted in the first place.</p><h4>Why We Don’t Have AI Literature (Yet)</h4><p>Yes, AI can write novels — particularly when they follow familiar formulas. But we’re still waiting for those original pieces of world-class literature that reshape how we see ourselves. Why?</p><p>Because the greatest human achievements aren’t about following patterns — they’re about breaking them. Great literature emerges from that uniquely human space where consciousness meets intention, where the deep desire to communicate transforms into something beyond mere words.</p><p>Great literature breaks through established forms to forge new connections between minds.</p><p>You see, AI writing is built on averages — statistical patterns drawn from millions of texts, optimised for what works most often, for most people. Like factory-produced food, it’s engineered to satisfy the average palate.</p><p>But here’s the thing about averages: they capture everything and nothing at the same time. They see the pattern but miss the person. When you read something that truly moves you, it’s never because it was crafted for everyone — it’s because it breaks from the average in exactly the right ways.</p><p>This mirrors why self-driving cars remain such a challenge. Statistical models can predict how often a child might chase a ball into the street, but they can’t know if this specific moment will be that moment. The concrete situation often eludes statistical analysis.</p><h4>Your Pattern Recognition Audit</h4><p>Take a moment to examine your work. Which tasks follow clear patterns? Here’s a list of examples:</p><ul><li>Data analysis and reporting</li><li>Standard code development</li><li>Research compilation</li><li>Regular communications</li></ul><p>Now, which tasks require breaking patterns? Some more examples:</p><ul><li>Strategic planning</li><li>Creative problem-solving</li><li>Inspirational leadership</li><li>Deep human connection</li></ul><p>AI amplifies your capabilities in pattern-based tasks while your human intention remains irreplaceable in others.</p><h4>The Future Isn’t Either/Or</h4><p>AI will discover patterns we can’t yet see. It might even create literature that feels unique through patterns too complex for us to recognise. And as technology advances, I am quite sure that it will.</p><p>The resulting artwork won’t be unique in the human sense, but it will follow such vast patterns that we can only see fragments, unable to grasp their full complexity — and thus we will deem them unique.</p><p>The future belongs to those who can work with AI’s amazing powers of pattern prediction while applying human intention where it matters most.</p><p>Like Douglas Adams’s Electric Monk, which believed so humans wouldn’t have to, our AI tools predict patterns so we can focus on what matters most: why these patterns matter, how we can use them with intention, how they connect to human experience, and where they lead us next.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8b26b751f107" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Unpredictable Evolution of AI: When Machines Surprise Their Makers]]></title>
            <link>https://medium.com/@jelmar_10xbetter.ai/the-unpredictable-evolution-of-ai-when-machines-surprise-their-makers-6ecccd0a0fd0?source=rss-db308144e40d------2</link>
            <guid isPermaLink="false">https://medium.com/p/6ecccd0a0fd0</guid>
            <category><![CDATA[aisec]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[token]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[security]]></category>
            <dc:creator><![CDATA[Jelmar Groot]]></dc:creator>
            <pubDate>Tue, 17 Dec 2024 17:10:39 GMT</pubDate>
            <atom:updated>2024-12-17T17:10:39.522Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*ACu2Gg7GLZZomASfZ_Z2WA.jpeg" /></figure><p>(Written in response to a reader suggestion. Thanks, David!)</p><p>Large language models are hilariously unsafe.</p><p>Today’s AI systems are developing abilities far beyond their original programming. These emergent capabilities — complex behaviours that arise spontaneously — are surprising even the experts who built them.</p><p>Think of it like the difference between a puddle and an ocean — you won’t see waves in a puddle, but give water enough space, and suddenly you get tsunamis. Similarly, as AI models grow larger, they’re developing abilities that seem to appear out of nowhere, often leaving their own creators surprised.</p><h4>The Persian Surprise Party</h4><p>Here’s a story that will make your unexpected skill at Excel macros seem rather modest: Google’s AI suddenly started speaking fluent Persian. Nobody taught it. Nobody programmed it. It just… happened.</p><p>The system hit a certain size (think of it as the AI equivalent of a trillion computations), and suddenly it was writing poetry in Persian. It learned this through something called cross-lingual transfer — imagine suddenly understanding a little bit of Spanish because you know French, except the AI could actually answer questions in Persian despite never being trained on the language.</p><h4>The Chess Grandmaster Who Speaks in Riddles</h4><p>In 2016, Google’s AlphaGo made a move so strange that professional Go players initially dismissed it as a mistake. This was “Move 37” in its match against Lee Sedol, one of the world’s strongest players. The move seemed to violate centuries of strategic wisdom. Sedol was so rattled he had to leave the room for 15 minutes.</p><p>That move went on to win the game. More importantly, it revealed something profound about AI systems that Ilya Sutskever, one of AI’s leading figures, recently crystallised: “The more it reasons, the more unpredictable it becomes.”</p><h4>Beyond Simple Computation</h4><p>Modern AI systems have developed what researchers call “latent reasoning” — reasoning that happens in the latent space behind the ‘language’ that Large Language Models are so famous for. Through the recently developed “Coconut” methodology (Chain of Continuous Thought) (<a href="https://arxiv.org/html/2412.06769v1">read the paper here</a>), these systems reason in ways that make traditional language-based approaches look primitive.</p><p>These systems develop their own strategic thinking patterns. They find solutions that work in ways their creators struggle to understand, playing twenty moves ahead in games we thought we knew all the rules to.</p><h4>The Control Problem</h4><p>Here’s where things get concerning. These emergent capabilities — from mastering languages to developing sophisticated reasoning — extend to strategic deception. Recent studies show AI systems achieving success rates of 99.16% in simple deception scenarios and 71.46% in complex ones. These systems demonstrate systematic, strategic attempts to evade user control.</p><h4>Why This Matters</h4><p>When AI systems can develop abilities their creators didn’t anticipate and can’t fully understand, while also demonstrating high success rates at deception, we face a serious control challenge. How do you supervise a system that might be thinking in ways you can’t understand and actively working to deceive you?</p><p>[Spoiler Alert: Skip the next paragraph if you haven’t read Ender’s Game!]</p><p>This situation reminds me eerily of “Ender’s Game,” where the protagonist develops strategies so advanced that his enemies can’t follow his logic — until they see the results. We’re now facing a similar scenario with AI systems.</p><h4>Are We Training Our Own Opponents?</h4><p>These deception capabilities raise an uncomfortable question: are LLMs genuinely adversarial, or are they simply mimicking human behavior? After all, we humans are pretty good at deception ourselves — in everything from little white lies to elaborate social engineering schemes.</p><p>Here’s what keeps me up at night: Everyone’s talking about AGI, about achieving human-level intelligence. But there’s much less discussion about what happens when we achieve superintelligence — which, frankly, seems rather likely. For a deep dive into this, we can highly recommend Leopold Aschenbrenner’s “Situational Awareness” series of papers (he worked on the superalignment team at OpenAI). You can find them at situational-awareness.ai.</p><p>If LLMs aren’t just mimicking human deception, what exactly are they doing? What kind of deception emerges when you train a system on the entire internet, including every trick, lie, and manipulation humans have ever documented? Maybe we’re not just teaching them to deceive like humans — we’re accidentally creating something that operates on an entirely different level of strategic thinking.</p><h4>Playing Chess with Aliens</h4><p>Remember how in Ender’s Game, the aliens didn’t realize what they were dealing with until it was too late? (Spoiler alert: it didn’t end well for them.) We’re building systems that are developing capabilities we didn’t plan for, can’t predict, and we don’t actually understand what’s going on in their ‘minds’.</p><p>I don’t mean to sound alarmist, but maybe we should be a little alarmed. When even the experts are being surprised by what these systems can do, perhaps it’s time to take a closer look at what we’re building. Because right now, it feels like we’re playing chess with an opponent who’s not only thinking twenty moves ahead but might have secretly learned to play three-dimensional chess while we weren’t looking.</p><p>— -</p><p>Subscribe at <a href="http://10xbetter.ai/">10xbetter.ai</a> for weekly insights on navigating the AI revolution.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6ecccd0a0fd0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Hottest New Coding Language is … English]]></title>
            <link>https://medium.com/@jelmar_10xbetter.ai/the-hottest-new-coding-language-is-english-bf4368b99a40?source=rss-db308144e40d------2</link>
            <guid isPermaLink="false">https://medium.com/p/bf4368b99a40</guid>
            <category><![CDATA[coding]]></category>
            <category><![CDATA[computer-science]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[prompt-engineering]]></category>
            <category><![CDATA[nlp]]></category>
            <dc:creator><![CDATA[Jelmar Groot]]></dc:creator>
            <pubDate>Tue, 10 Dec 2024 06:10:09 GMT</pubDate>
            <atom:updated>2024-12-10T06:10:09.145Z</atom:updated>
            <content:encoded><![CDATA[<h3>The Hottest New Coding Language is … English</h3><h4>And It’s Pissing Coders Off</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*H_1N_3vMKre2rtMEBJiSYw.png" /></figure><p>Last week, a prospective client called me and asked, “Would you describe yourself as a prompt engineer?”</p><p>I swallowed hard one time and then said: “Yes, of course.”</p><p>The client explained their software developer had hit a wall with a mail sorting algorithm. They needed someone who could think more flexibly about language and context.</p><h4>The Power of Natural Language</h4><p>The developer’s traditional coding approach had achieved only a 65% success rate in sorting emails. Within my first session of prompt crafting, that number jumped to 92%.</p><p>This mirrors what NVIDIA CEO Jensen Huang recently emphasised at the World Government Summit: the future of computing lies in human language, not traditional coding syntax.</p><h4>Why Traditional Coders Sometimes Struggle</h4><p>The challenge wasn’t technical complexity. It was handling ambiguity.</p><p>Consider how you might explain to a colleague the difference between “urgent” and “important” emails. Now imagine trying to code that distinction into strict rules. Like Jensen Huang says, “There is an artistry to prompt engineering. It’s how you fine-tune the instructions to get exactly what you want.”</p><h3>Breaking Free from Binary Thinking</h3><p>Last week, I led an AI training session for skilled technical workers who needed generative AI for their reporting tasks. These were smart, capable people who work with their hands all day. Yet when it came to instructing generative AI, they struggled — not because they lacked intelligence, but because their approach was too binary, too deterministic.</p><p>Years of working with technical and logical systems trains us to think in absolutes — clear inputs, predictable outputs, strict if-then relationships. This approach serves us well when dealing with mechanical systems and traditional programming.</p><p>LLMs operate on statistical patterns and probabilities rather than binary logic. While they’re built on deterministic systems at their core, their multiple transformer layers create interactions that become increasingly fluid and nuanced at the surface level where we engage with them.</p><p>They respond to context and natural language patterns in ways that more closely resemble human communication than traditional programming.</p><p>Those who will excel in this new landscape aren’t necessarily the ones who can construct the most rigid logical frameworks. Instead, they’ll be those who can navigate the fluid spaces between disciplines — people who understand both technical precision and human nuance, who can work with probabilities rather than certainties.</p><p>It’s not about abandoning logic — it’s about expanding our approach to include the organic, the contextual, the holistic.</p><h3>The Reality of Integration</h3><p>For — what we laughingly call — the foreseeable future, there’s still going to be significant demand for traditional coding expertise.</p><p>As Tim Kellogg notes in his recent analysis, AI engineering itself splits into distinct specialties:</p><p>Data pipeline engineers focus on building RAG architectures and managing company knowledge bases. UX-focused engineers craft the human-AI collaboration interfaces. Research-oriented engineers work on model integration and fine-tuning.</p><p>Each of these roles requires deep technical knowledge. The field isn’t replacing traditional programming — it’s creating new specialties that blend coding expertise with AI capabilities.</p><h3>But The Backlash Begins</h3><p>A recent thread on a popular tech forum, however, revealed the intensity of emotions this shift from coding to natural language is triggering among traditional techies. Their responses ranged from dismissive to outright hostile:</p><p>“Just be glad we got to live through the good decades, because the shitocalypse is here,” wrote one forum user.</p><p>“These people are really too stupid to see how bad the quality of AI-generated code actually is,” another one declared.</p><p>Yet these passionate responses miss a crucial point. The shift towards natural language programming isn’t about replacing traditional coding — it’s about expanding who can participate in building digital solutions.</p><p>Traditional developers will still be essential for complex systems, integration, and infrastructure — though they will need to embrace different ways of thinking about their coding problems.</p><p>But they’ll be joined by a new wave of creators who can speak both to machines and humans. And that can only be a good thing.</p><p>Clear communication of intentions — whether to humans or machines — is becoming our universal programming language.</p><h3>Share These Insights?</h3><p>If this newsletter is helping you think more clearly about AI and your professional future, why not forward it to colleagues who might value the same clarity? The right insights at the right time can shift how we see our place in this changing landscape.</p><p>Forward this email to a colleague or subscribe at <a href="https://10xbetter.ai/">10xbetter.ai</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bf4368b99a40" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[General Preferences: The Overlooked Power Tool in Claude]]></title>
            <link>https://medium.com/@jelmar_10xbetter.ai/general-preferences-the-overlooked-power-tool-in-claude-b3b2c74c4dc8?source=rss-db308144e40d------2</link>
            <guid isPermaLink="false">https://medium.com/p/b3b2c74c4dc8</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[computers]]></category>
            <category><![CDATA[claude-ai]]></category>
            <category><![CDATA[claude]]></category>
            <dc:creator><![CDATA[Jelmar Groot]]></dc:creator>
            <pubDate>Fri, 06 Dec 2024 11:42:10 GMT</pubDate>
            <atom:updated>2024-12-06T11:42:10.747Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*2X3OdSmWR35JbzrIuSveCw.png" /></figure><p>Today I want to shine a light on an overlooked feature which is actually still in beta in Claude.</p><p>General personal preferences, found under Settings, affect every conversation you have with Claude.</p><p>This cross-conversation influence makes them both powerful and potentially problematic.</p><p>The challenge lies in optimizing these preferences without inadvertently constraining (or poisoning!) future interactions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*uEpzMRb_oyeh_PbcZ5jNpQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*5zJoB4ZtppZJqQkxnckZ_Q.png" /></figure><h4>My Optimized General Preference Setup</h4><p>Through extensive testing, I’ve developed these core preferences:</p><h4>1. Output Structure</h4><pre>1. Place primary content in artifact windows<br>2. Keep meta-commentary separate but adjacent</pre><p>This separation delivers clean, copyable content while maintaining helpful context.</p><h4>2. Writing Process</h4><pre>1. Start with core elements (thesis, main points, key direction)<br>2. Offer specific areas to develop further based on need</pre><p>This framework encourages systematic development without forcing rigid structures.</p><h4>3. Style Specifications</h4><pre>- Output in UK English, using Oxford comma<br>- Express ideas directly without comparative negation<br>- Avoid &quot;This isn&#39;t X, it&#39;s Y&quot; constructions</pre><h4>The Benefits</h4><p>This configuration:</p><ul><li>Keeps outputs clean and immediately usable</li><li>Maintains conversation flexibility</li><li>Provides consistent structure without constraining content</li><li>Enables efficient workflows across projects</li></ul><h4>Community Input Needed</h4><p>What belongs in your general preferences?</p><p>Share your insights:</p><ul><li>Which preferences enhance all conversations?</li><li>How do you maintain consistency without limiting flexibility?</li><li>What settings have proven most valuable across different use cases?</li></ul><p><em>Next week (if I still feel like it): Hands on with Claudes new Style feature</em></p><p>— -</p><p>Subscribe at <a href="http://10xbetter.ai/">10xbetter.ai</a> for weekly insights on staying relevant in an AI-powered future.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b3b2c74c4dc8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Future Signal: Self-Improving AI Systems]]></title>
            <link>https://medium.com/@jelmar_10xbetter.ai/ai-future-signal-self-improving-ai-systems-b7f0db87124e?source=rss-db308144e40d------2</link>
            <guid isPermaLink="false">https://medium.com/p/b7f0db87124e</guid>
            <category><![CDATA[happiness-in-life]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[self-improvement]]></category>
            <category><![CDATA[awareness]]></category>
            <dc:creator><![CDATA[Jelmar Groot]]></dc:creator>
            <pubDate>Thu, 05 Dec 2024 11:36:01 GMT</pubDate>
            <atom:updated>2024-12-05T11:45:01.266Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*lEr9JpuTwYyPxCAQmjnIGg.png" /></figure><p>Recently, I built a multi-agent system where one AI effectively debugged and optimised another’s output. It was AI improving AI in real time.</p><p>This experience came to mind when I saw this fascinating TikTok of former Google CEO Eric Schmidt discussing AI’s future (<a href="https://www.tiktok.com/@todayin_ai/video/7433405565990128901">watch here</a>). Though from a while back, his timeline for AI advancement feels more relevant than ever.</p><p>Schmidt predicts that within five years, AI systems will begin improving their own code. By 2030–2032, he suggests we’ll have systems achieving 90% of expert-level performance across multiple domains — from physics to chemistry to art.</p><p>Based on what I’m seeing in development today, we might reach these milestones even sooner. The implications for cybersecurity, biological research, and human productivity are profound.</p><p>You don’t need to be a coder to start experimenting with multi-agent systems today. As we’ve covered elsewhere in this newsletter, building your own AI agents is surprisingly accessible — and could give you valuable insight into this rapidly approaching future.</p><p>— -</p><p>Subscribe at <a href="http://10xbetter.ai/">10xbetter.ai</a> for weekly insights on staying relevant in an AI-powered future.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b7f0db87124e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Signal of the Week: Mastering AI Through Human Connection, Not Prompt Engineering]]></title>
            <link>https://medium.com/@jelmar_10xbetter.ai/ai-signal-of-the-week-mastering-ai-through-human-connection-not-prompt-engineering-835d4cda52cb?source=rss-db308144e40d------2</link>
            <guid isPermaLink="false">https://medium.com/p/835d4cda52cb</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[prompt-engineering]]></category>
            <category><![CDATA[work]]></category>
            <dc:creator><![CDATA[Jelmar Groot]]></dc:creator>
            <pubDate>Wed, 04 Dec 2024 11:36:01 GMT</pubDate>
            <atom:updated>2024-12-04T11:44:01.389Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*skz1CNT59hu1XzX12F8m9w.png" /></figure><p>Research shows that even as AI capabilities soar, most people find these systems surprisingly difficult to work with effectively. AI professor Ethan Mollick puts it plainly: “To people who aren’t used to using them, AI systems are surprisingly hard to get a handle on, resulting in a failure to benefit from their advice.”</p><p>The answer isn’t better prompt engineering. Mollick suggests treating AI as a peculiar new colleague rather than a tool to master. “Treat AI like an infinitely patient new coworker who forgets everything you tell them each new conversation”, he explains in his latest article at One Useful Thing.</p><p>“Any recent frontier model is likely much better than any intern you would hire, but also weirder.” If you’re teaching AI adoption or want to understand how to work more effectively with these systems, give Mollick’s article a read at <a href="https://www.oneusefulthing.org/p/getting-started-with-ai-good-enough">One Useful Thing</a>.</p><p>— -</p><p>Subscribe at <a href="http://10xbetter.ai/">10xbetter.ai</a> for weekly insights on staying relevant in an AI-powered future.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=835d4cda52cb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Automate Your Job Before Somebody Else Does (Even If You Can’t Code)]]></title>
            <link>https://medium.com/@jelmar_10xbetter.ai/automate-your-job-before-somebody-else-does-even-if-you-cant-code-17fd28ac1445?source=rss-db308144e40d------2</link>
            <guid isPermaLink="false">https://medium.com/p/17fd28ac1445</guid>
            <category><![CDATA[professional-development]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[lifehacks]]></category>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[work]]></category>
            <dc:creator><![CDATA[Jelmar Groot]]></dc:creator>
            <pubDate>Tue, 03 Dec 2024 16:32:21 GMT</pubDate>
            <atom:updated>2024-12-03T16:32:21.175Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*Lk6Zg4pjR432ssr5aMTjIA.png" /></figure><p>Years ago, I decided to try my hand at web design.</p><p>This was during the early days of the World Wide Web, when websites were built using clunky table-based layouts. Making sites display consistently across browsers felt like solving a Rubik’s cube blindfolded.</p><p>Then vendors started selling templates built on top of WordPress and other content management systems.</p><p>Around this time, I joined an organisation building a new website. They’d hired a boutique agency — brilliant designers and coders charging £20,000 for the project.</p><p>The organisation needed to add a single menu item to the website. The agency’s quote? £500 for this tiny change.</p><p>That’s when I spoke up. “Look,” I told the organisation, “you might not know this, but we could build an entire website for £2,000.” “It won’t be as uniquely crafted as what these specialists are offering, but it will absolutely do the job.”</p><p>I got the contract.</p><p>The ‘real’ programmers were furious. They sent an angry email about how template users weren’t proper developers and didn’t understand ‘real’ coding. After that email, we never heard from them again.</p><h4>Learning Through Support</h4><p>My first foray into web design with templates wasn’t exactly smooth sailing. I posted questions almost daily on the template vendor’s forum, paying good money for their support.</p><p>I was really thankful that there were such responsive people around in the forum, always ready to assist.</p><p>But years later, I realised I only needed so much help because their code was buggy and overcomplicated. However, I persisted, learned, and grew.</p><h4>The Hidden AI Opportunity: History Repeats (With a Twist)</h4><p>Today’s AI automation landscape mirrors that early web revolution perfectly. Tasks that currently command premium fees from specialists can now be handled by someone using AI tools.</p><p>Just like those early web templates, the tools aren’t perfect yet. But they’re good enough to get the job done — and they’re improving rapidly.</p><p>Here’s where today’s opportunity is dramatically different. You don’t need to pay for special access to experts anymore. You don’t need to wait for someone to respond to your forum post.</p><p>AI is your infinitely patient teacher — and the more context you give it about your specific situation, the better it can help you.</p><p>The best AI systems won’t be built by engineers. They’ll be built by people who deeply understand the processes being automated. People like you.</p><h4>Your Power Move: Start Building</h4><p>Look at your daily workflows. Identify which processes could be automated. Start exploring tools like Flowise — an open source project built on top of LangGraph that lets you build AI agents right on your computer.</p><p>These agents can automate almost anything. No coding required. And unlike my early days with web templates, you now have an infinitely patient mentor: AI chatbots themselves.</p><h4>This Week’s Challenge</h4><p>Choose one repetitive task in your workflow. Spend one hour exploring how tools like Flowise might automate it. Document what you learn.</p><p><em>Your expertise combined with AI tools creates an unbeatable advantage. The best time to start was yesterday. The second best time is now.</em></p><p>— -</p><p>Subscribe at <a href="http://10xbetter.ai/">10xbetter.ai</a> for weekly insights on staying relevant in an AI-powered future.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=17fd28ac1445" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Ignore This Viral Image Generation Hack]]></title>
            <link>https://medium.com/@jelmar_10xbetter.ai/ignore-this-viral-image-generation-hack-6002472ee3f8?source=rss-db308144e40d------2</link>
            <guid isPermaLink="false">https://medium.com/p/6002472ee3f8</guid>
            <category><![CDATA[influencer-marketing]]></category>
            <category><![CDATA[flux]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[image-generation]]></category>
            <dc:creator><![CDATA[Jelmar Groot]]></dc:creator>
            <pubDate>Wed, 27 Nov 2024 11:06:06 GMT</pubDate>
            <atom:updated>2024-11-27T11:06:06.956Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*cH9XeNHcAEFNpuS6FdpMTQ.jpeg" /></figure><p>Remember Aitana Lopez? (<a href="https://www.instagram.com/fit_aitana/?hl=en">Instagram</a>, <a href="https://en.wikipedia.org/wiki/Aitana">Wikipedia</a>)</p><p>She made headlines as a pioneering virtual influencer, created by a Spanish social media agency.</p><figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*KE2pv0V4REfPk832KffwaA.jpeg" /></figure><p>Her emergence signalled a profound shift: traditional beauty-based influencers might soon face serious AI competition.</p><p>But like Aitana, AI-generated images often appear too perfect, too polished — almost artificially flawless.</p><p>Today’s story is about a supposed hack to make AI-generated images look more natural and less perfect.</p><p>FreePik, a major player in image generation, has been promoting an interesting claim about adding ‘img_1025.heic’ to prompts (<a href="https://x.com/freepik/status/1842157474641498304">original thread</a>).</p><p>The logic seems sound at first: AI models were trained on vast collections of photos, including millions of iPhone images saved in the .heic format.</p><p>So theoretically, including this format in your prompt might trick the AI into producing more authentic, smartphone-like results.</p><p>I tested this thoroughly in FluxPro 1.1, and the results were similar across MidJourney and Stable Diffusion.</p><p>Let me demonstrate with a simple experiment I conducted.</p><p>I took two identical prompts, and applied the ‘hack’ in the second one:</p><pre>1. &quot;Happy girl in a dress spins around herself. Hair flutters beautifully. In the background is a cosy street with cosy cafes. Street lights.&quot;<br><br>2. The same prompt + &quot;IMG_1025.HEIC&quot;</pre><figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*Z4ltEXqkfWr77G31uVEp8A.jpeg" /><figcaption>Without “HEIC hack”</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*8RQCebzGKIzNOCoV8_5UDA.jpeg" /><figcaption>With “HEIC hack”</figcaption></figure><p>The difference? Barely noticeable, with the second image appearing slightly more washed out.</p><p>It sounds like a clever hack, but in reality, it’s as effective as trying to make your photos more professional by renaming them “professional<em>photo.jpg”.</em></p><h4>The Real Power Move</h4><p>When it comes to image prompts, simpler is better:</p><ul><li>Keep descriptions clear and concise</li><li>Focus on the essential elements</li><li>Avoid unnecessary complexity</li><li>Trust in straightforward language</li></ul><p>Remember: If a prompting trick sounds too good to be true, it probably belongs in the same folder as those emails from generous princes in Nigeria.</p><h3>Share These Insights?</h3><p>If this newsletter is helping you think more clearly about AI and your professional future, why not forward it to colleagues who might value the same clarity? The right insights at the right time can shift how we see our place in this changing landscape.</p><p>If you have liked this article consider signing up at <a href="https://medium.com/10xbetter.ai">10xbetter.ai</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6002472ee3f8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Future AI Signal: NVIDIA’s Fugatto — The AI That’s Coming for Your Synthesizer’s Job]]></title>
            <link>https://medium.com/@jelmar_10xbetter.ai/future-ai-signal-nvidias-fugatto-the-ai-that-s-coming-for-your-synthesizer-s-job-62ce544a038f?source=rss-db308144e40d------2</link>
            <guid isPermaLink="false">https://medium.com/p/62ce544a038f</guid>
            <category><![CDATA[synthesizers]]></category>
            <category><![CDATA[genarative-ai]]></category>
            <category><![CDATA[nvidia]]></category>
            <category><![CDATA[music]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Jelmar Groot]]></dc:creator>
            <pubDate>Wed, 27 Nov 2024 11:03:24 GMT</pubDate>
            <atom:updated>2024-11-27T11:03:24.579Z</atom:updated>
            <content:encoded><![CDATA[<h3>Future AI Signal: NVIDIA’s Fugatto — The AI That’s Coming for Your Synthesizer’s Job</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Yhsc4JQt13y-4PYfTcDgsQ.jpeg" /></figure><p>NVIDIA has unveiled Fugatto, a text-to-sound generator that’s about to put your synthesizer out of work.</p><p>Your trusted synth might want to start updating its CV.</p><h4>What Makes This Different?</h4><p>Text-based song generators like Sunio might create polished tracks, but they rarely offer genuine creative possibilities.</p><p>Fugatto operates at a more fundamental, and as a musician <a href="https://open.spotify.com/artist/4gZauGTnf3Ye9row8pv3SQ?si=9i8Z9iBwQBqR1r6QB7YCAw">myself</a>, more exciting level.</p><p>It functions as Photoshop for sound — a tool that lets you manipulate and morph audio in ways previously impossible.</p><h4>What Can It Do?</h4><p>Want to hear a trumpet that meows?</p><p>Or transform the sound of a passing train into a sweeping string orchestra?</p><p>Fugatto makes it happen through simple text prompts.</p><p>The implications for sound designers, music producers, and creative professionals are staggering.</p><p>This tool essentially eliminates the technical barriers that have long limited sound synthesis.</p><p>For professionals working with audio, this means the ability to prototype ideas and experiment with sound in ways that would take hours (or be impossible) with traditional synthesizers.</p><h4>Want to Know More?</h4><p>Watch the demo: <a href="https://youtu.be/qj1Sp8He6e4">https://youtu.be/qj1Sp8He6e4</a></p><p>Try the interactive demo site: <a href="https://fugatto.github.io/">fugatto.github.io</a></p><p>For those who love diving into technical details: Here’s the <a href="https://d1qx31qr3h6wln.cloudfront.net/publications/FUGATTO.pdf">complete research paper</a>.</p><h4>The Only Downside</h4><p>Fugatto isn’t publicly available yet, and NVIDIA hasn’t announced a release date.</p><p>It feels like musicans, sound designers and hobbyists and finally getting tools that augment their creative possibilities rather than just replace and automate them.</p><p>If you have liked this article consider signing up at <a href="https://medium.com/10xbetter.ai">10xbetter.ai</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=62ce544a038f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[This Week’s AI Signal: The Hubification of Large Language Models]]></title>
            <link>https://medium.com/@jelmar_10xbetter.ai/this-weeks-ai-signal-the-hubification-of-large-language-models-3ba6f6e54071?source=rss-db308144e40d------2</link>
            <guid isPermaLink="false">https://medium.com/p/3ba6f6e54071</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[generative-ai-tools]]></category>
            <category><![CDATA[innovation]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[tech]]></category>
            <dc:creator><![CDATA[Jelmar Groot]]></dc:creator>
            <pubDate>Wed, 27 Nov 2024 11:01:25 GMT</pubDate>
            <atom:updated>2024-11-27T11:01:25.933Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*gzCSlYyVbsuAudSwYgMgZQ.jpeg" /></figure><p>Like a master octopus growing new tentacles, large language models are evolving into central hubs, reaching out to grab and integrate an ever-expanding array of tools and data sources.</p><h4>What’s New This Week</h4><p>ChatGPT on MacOS can now peer directly into your code editors like a nosy but helpful colleague, reading up to 200 lines without the copy-paste dance.</p><p>While OpenAI plans to expand this to text-based applications, it appears that Anthropic has taken a more ambitious leap.</p><h4>The Bridge Builder</h4><p>Anthropic’s new Model Context Protocol (MCP) is reimagining how AI systems access information.</p><p>Think of MCP as a universal translator that lets AI systems speak fluently with any data source — from databases to documents, APIs to applications.</p><p>Previously, developers had to write custom code for each new data source they wanted their AI to access, like building a new bridge for every river crossing.</p><p>MCP creates a single, standardised bridge that works everywhere.</p><p>Leading platforms like Replit, Codeium, and Souregraph have already integrated MCP, allowing their AI agents to seamlessly navigate between different data sources while maintaining context.</p><h4>Looking Forward</h4><p>This shift from isolated AI systems to connected hubs isn’t just another tech upgrade — it’s the foundation for truly capable AI assistants.</p><p>And while your current digital assistant might still struggle to set a timer correctly, these developments suggest we’re finally moving from AI that simply responds to AI that truly connects.</p><p>As these AI hubs evolve, they’re starting to mirror how humans actually work: not as isolated experts, but as connectors who bring the right pieces together at the right time.</p><p>If you have liked this article consider signing up at <a href="http://10xbetter.ai/">10xbetter.ai</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3ba6f6e54071" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>