<?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[Stupid Human - Medium]]></title>
        <description><![CDATA[We’re AI scientists, creatives &amp; strategists that find innovative and creative applications of AI to solve problems in organisations and society. From reducing national skin-cancer mortality to scaling content operations 900× we deliver impact through the creative use of AI. - Medium]]></description>
        <link>https://medium.com/stupid-human?source=rss----d02530b3b98f---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>Stupid Human - Medium</title>
            <link>https://medium.com/stupid-human?source=rss----d02530b3b98f---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 03:48:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/stupid-human" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Price You Pay for Letting AI Shop for You]]></title>
            <link>https://medium.com/stupid-human/the-price-you-pay-for-letting-ai-shop-for-you-3b50281b8859?source=rss----d02530b3b98f---4</link>
            <guid isPermaLink="false">https://medium.com/p/3b50281b8859</guid>
            <category><![CDATA[e-commerce-business]]></category>
            <category><![CDATA[generative-ai-market]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[llm]]></category>
            <dc:creator><![CDATA[Jasmine Rienecker]]></dc:creator>
            <pubDate>Mon, 09 Feb 2026 10:43:31 GMT</pubDate>
            <atom:updated>2026-02-09T10:43:29.835Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*qEfuHWPIpJzYHNN7" /></figure><p>As AI agents take over searching, negotiating, and transacting in markets, outcomes are shaped less by consumer preference and more by agent design. This shift quietly determines who gets better prices, better options, and better outcomes.</p><p>This shift is easy to miss because it arrives wrapped in the familiar promise of efficiency. AI agents can now autonomously search, compare, negotiate, and complete transactions across platforms, collapsing complex, multi-step consumer journeys into unified flows. Tasks, like moving house, that once required dozens of coordinated decisions can be handled end-to-end by AI acting on the user’s behalf. Consultancies have also been quick to frame this as an unambiguous win. McKinsey projects that autonomous AI agents could orchestrate over $1 trillion in U.S. retail revenue by 2030, and $3–5 trillion globally. This is a transformation comparable to the rise of e-commerce but unfolding faster because the agents can operate on existing digital infrastructure.</p><p>Yet this efficiency-centred framing misses a deeper shift. A recent Stanford study showed that agents pursuing the same objective consistently achieved different prices; not due to randomness, but because the agents search, negotiate, and settle in different ways. This means access to more capable AI models could start translating to better deals, reinforcing economic disparities among users. At the same time, limitations in numerical reasoning and failures in instruction-following routinely caused the agents to violate user constraints, exceeding budget limits and overpaying relative to retail prices.</p><p>Beneath the surface of “autonomous decision-making” lie a few recurring mechanisms that consistently shape how agents behave. Understanding these patterns is essential to predicting how agentic markets will actually function.</p><h3>1. Decision Truncation</h3><p>Consider a buyer agent tasked with purchasing a laptop under a fixed budget. The agent queries the market and evaluates offers as they arrive, assessing conditions such as price, delivery time, and reliability. Now consider two sellers: one responds almost immediately with a decent but overpriced offer, while another responds more slowly with a clearly superior deal. In many agent-mediated systems, the first seller wins.</p><p>This pattern appears repeatedly in the <em>Magentic Marketplace </em>simulation, where agents that made the first proposal were chosen between 60–100% of the time depending on the model, even when their offers were objectively worse. By the third respondent, selection rates had decreased to near-zero. GPT-4o and Sonnet-4.5 showed the most extreme behaviour, selecting the first proposal 100% of the time, meaning these agents never even waited to compare alternatives.</p><p>This is not manipulation or platform bias, but the way agent decision rules collapse search depth, response speed, and acceptance thresholds into a single stopping condition. This finding suggest agents are not genuinely comparing options but stopping once an offer is “good enough.”</p><p>The result is that speed will start to become more important than quality. Sellers will compete to respond first rather than to offer optimal value, producing fast but systematically suboptimal outcomes.</p><h3>2. Adaptation and Repeated Play</h3><p>These dynamics intensify once agents and sellers interact repeatedly. Suppose a travel-booking agent consistently accepts hotel offers within a narrow price band. Over time, the hotel’s agent could learn this pattern, stop offering deeper discounts up front, and instead converge on prices just inside the travel-booking agent’s acceptance threshold. Transactions continue, metrics look healthy and the agent continues to “optimise”, but the effective market price has shifted upward.</p><p>This isn’t just a thought experiment. In Germany’s retail gasoline market, algorithmic pricing adoption increased profit margins by about 9% on average in non-monopoly markets. The effects were much larger in highly concentrated local duopolies: when both nearby stations adopted algorithmic pricing, margins rose by around 28%. Each firm was simply optimising its own pricing, but together the algorithms ended up sustaining higher prices.</p><p>Similar dynamics are well documented in automated negotiation more broadly. Learning-based agents such as ANEGMA adapt their strategies based on previous interactions, while their counterparts do the same. Over time, these systems reliably converge to stable equilibria that are locally optimal for the agents but often systematically disadvantage buyers and fall short of broader efficiency or fairness goals. Crucially, because welfare and fairness are not explicit objectives in agentic commerce, these outcomes are not treated as failures; they are simply the natural result of independently optimising agents learning how to coexist.</p><h3>3. Market Legibility</h3><p>Agents can only optimise over the information they can process. In practice, this means that agents prioritise structured, machine-readable attributes such as price, delivery time and standardised product fields, while qualitative factors like customer support quality, brand trust, or contextual fit are ignored unless they are explicitly encoded.</p><p>Protocol design amplifies this effect. Fixed schemas and interaction patterns teach sellers which attributes matter (and which do not), and they optimise accordingly. Although digital commerce has long rewarded visibility and conformity to algorithmic structures (SEO, ranking systems do this today), the crucial difference is that traditional platforms rank options for humans. While exploration, hesitation, and override once allowed humans to correct or resist algorithmic bias, agentic agents execute on the user’s behalf, removing these mechanisms.</p><p>It would be a mistake to frame agentic commerce as an entirely new game. Like search engines and marketplaces, speed, visibility, and legibility still win. Firms that adapt best to the underlying technical structure gain an advantage, just as they did before. What has changed is that markets are no longer optimising for human comparison but for agent behaviour. Outcomes depend less on abstract intent alignment and more on how agents search, stop, learn, and interact. Two users with identical goals can experience systematically different outcomes because their agents play different strategic games.</p><p>McKinsey is right that agentic commerce reduces friction. But friction was not just waste; it was exploration, comparison, and diversity of choice. As agents become primary market participants, these qualities need to become matters of design. The next phase of agentic commerce will be defined not by whether agents are “correct,” but by whether markets are designed with a clear understanding of agent behaviour.</p><p><strong>Source Links:</strong></p><ul><li><a href="https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-agentic-commerce-opportunity-how-ai-agents-are-ushering-in-a-new-era-for-consumers-and-merchants?utm_source=chatgpt.com">Agentic commerce: How agents are ushering in a new era</a></li><li><a href="https://arxiv.org/abs/2506.00073">The Automated but Risky Game: Modeling and Benchmarking Agent-to-Agent Negotiations and Transactions in Consumer Markets</a></li><li><a href="https://www.journals.uchicago.edu/doi/abs/10.1086/726906?journalCode=jpe">Algorithmic Pricing and Competition: Empirical Evidence from the German Retail Gasoline Market</a></li><li><a href="https://link.springer.com/article/10.1007/s10458-021-09513-x">ANEGMA: an automated negotiation model for e-markets</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3b50281b8859" width="1" height="1" alt=""><hr><p><a href="https://medium.com/stupid-human/the-price-you-pay-for-letting-ai-shop-for-you-3b50281b8859">The Price You Pay for Letting AI Shop for You</a> was originally published in <a href="https://medium.com/stupid-human">Stupid Human</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The GEO Illusion: What Actually Drives AI Visibility]]></title>
            <link>https://medium.com/stupid-human/the-geo-illusion-what-actually-drives-ai-visibility-6c19dbd56f2b?source=rss----d02530b3b98f---4</link>
            <guid isPermaLink="false">https://medium.com/p/6c19dbd56f2b</guid>
            <category><![CDATA[ai-visibility]]></category>
            <category><![CDATA[large-language-models]]></category>
            <category><![CDATA[seo]]></category>
            <dc:creator><![CDATA[Jasmine Rienecker]]></dc:creator>
            <pubDate>Thu, 20 Nov 2025 10:04:02 GMT</pubDate>
            <atom:updated>2025-11-20T10:46:44.642Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*npe8G8zp4AGk4KpTJ5UQgg.jpeg" /></figure><p>AI assistants like ChatGPT, Gemini and Meta AI are rapidly becoming primary gatekeepers of brand discovery. When users turn to these systems for recommendations, being invisible in their responses means your brand risks disappearing from consumer awareness. Companies that have doubled down on traditional SEO and GEO tactics are finding fewer citations in AI-generated answers, while competitors with strong cross-domain authority show up more often, even when their own websites are technically weaker.</p><h3>The Limits of SEO and GEO Strategies</h3><p>This isn’t a failure of execution; it’s a mismatch in strategy. Both SEO and GEO target single-domain signals such as page structure and on-site content, while AI-driven systems prioritise multi-domain validation patterns. Analysis of citation patterns across major AI platforms reveals that claims often require either multiple independent domain confirmations or exceptionally high single-source trust scores before earning citation status.</p><h3>How AI Systems Really Choose Sources</h3><p>To understand this pattern, you need to follow how AI systems transform web content into authoritative answers.</p><p>AI systems begin in familiar territory: crawling, indexing, and parsing web content for use in both training and search databases. Traditional SEO hygiene still matters here as clean markup, proper headings, XML sitemaps and canonicals help ensure your content gets discovered and processed correctly.</p><h3>Surviving the Chunking Process</h3><p>The divergence from traditional search engines begins during content processing, where websites are chunked into digestible pieces. Here, GEO techniques briefly take centre stage, as signals like DOIs, clear attributions, and JSON-LD make a chunk easier for the system to parse and index. A short, well-tagged claim wrapped in JSON-LD is much more likely to survive the chunking process intact than the same claim buried in a long, unstructured page. This means GEO optimisation shouldn’t be overlooked, as it ensures your content maintains its structure and authority markers even when broken down into fragments for AI processing.</p><p>However, surviving chunking is only the first step. When users ask open-ended questions about brands or products, queries that trigger external web retrieval rather than relying solely on the model’s training, your content faces two more hurdles.</p><p>The system first matches the user’s query to the content chunks using a combination of lexical and semantic matching. Passing this filter gives a chunk the right to compete, but it doesn’t guarantee visibility.</p><p>The decisive stage is re-ranking: a multi-step process that turns hundreds of candidate chunks into the select few that will actually get cited. During re-ranking, the system clusters candidate passages by their source domain and runs domain-trust evaluation.</p><h3>Why Domain Trust Matters</h3><p>This evaluation looks beyond on-page SEO and GEO by considering factors like a publisher’s reputation in academic indexes, the quality and authority of inbound links, whether the content can be found in institutional archives, and how often respected outlets reference your domain. The system is scoring entire domains for credibility, not just individual pages for polish.</p><p>Next comes support aggregation, where the re-ranker counts how many distinct domains support the same claim and then weighs those counts against each domain’s trust score. Most systems only treat a claim as corroborated when it appears across multiple independent domains, or when it’s backed by a single source with exceptional authority.</p><p>This re-ranking stage creates a hard limit for site-focused strategies: publishing the same claim across ten pages of your website still counts as just one supporting voice. No amount of internal duplication boosts the number of independent confirmations the AI cares about.</p><h3>Building Cross-Domain Authority</h3><p>Breaking through this validation barrier requires a fundamental shift in approach. Rather than focusing solely on optimising your own digital properties, successful companies are building systematic programs that cultivate independent confirmations across authoritative domains.</p><h3>01 Monitoring</h3><p>Establish baseline visibility into the AI-generated answers by running industry-relevant questions that potential customers might ask. Documenting which companies, sources, and publications appear consistently in responses allows identifying trends into the types of sources being referenced and measuring your progress over time. The goal is not vanity metrics but creating a working map of the evidence AI systems actually use in your vertical.</p><h3>02 Source Intelligence</h3><p>Profile the authoritative sources that show up in your monitoring to understand what each domain values. Understanding these preferences, whether that be original datasets, short expert commentary or long-form research, is crucial for positioning your insights effectively and therefore gives you the best chance of being cited independently.</p><h3>03 Systematic Experimentation</h3><p>Design targeted experiments to establish thought leadership across identified authoritative sources. This involves creating content specifically designed to be referenced by independent publications rather than optimising your own website. This could be in the form of original research, proprietary frameworks that consultants and analysts might adopt, or educational resources that academic institutions could cite. For every experiment, measuring distinct domain references, assessing the trust quality of those domains, and tracking changes in AI citation patterns provides actionable insights for refinement.</p><h3>The New Definition of Visibility</h3><p>The fundamental shift here is to stop treating visibility as a property of your site alone and start treating it as a property of the network of independent, referenceable confirmations that AI re-rankers actually count.</p><p>As AI systems handle growing shares of business research, invisibility in AI answers increasingly means invisibility to prospects. Companies succeeding in AI visibility understand that optimisation and authority operate on different planes. They maintain technical excellence while building something more valuable: independent validation of their expertise across multiple domains.</p><p>These organisations focus on making their insights referenceable rather than just discoverable and cultivate mentions through genuine thought leadership. SEO and GEO get you into the candidate pool and help you survive initial processing, but independent authority determines whether AI systems trust you enough to quote.</p><p>In this new landscape, the winners won’t be those with the most optimised websites, they’ll be the organisations whose expertise is so well-established across independent sources that AI systems naturally recognise them as trusted authorities.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6c19dbd56f2b" width="1" height="1" alt=""><hr><p><a href="https://medium.com/stupid-human/the-geo-illusion-what-actually-drives-ai-visibility-6c19dbd56f2b">The GEO Illusion: What Actually Drives AI Visibility</a> was originally published in <a href="https://medium.com/stupid-human">Stupid Human</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Competitive Advantage Through Regulation: Europe’s Vision for Responsible AI]]></title>
            <link>https://medium.com/stupid-human/competitive-advantage-through-regulation-europes-vision-for-responsible-ai-6722e48853ba?source=rss----d02530b3b98f---4</link>
            <guid isPermaLink="false">https://medium.com/p/6722e48853ba</guid>
            <category><![CDATA[ai-regulation]]></category>
            <category><![CDATA[ai-ethics]]></category>
            <category><![CDATA[ai-compliance]]></category>
            <category><![CDATA[eu-ai-act]]></category>
            <dc:creator><![CDATA[Jasmine Rienecker]]></dc:creator>
            <pubDate>Thu, 20 Nov 2025 09:57:31 GMT</pubDate>
            <atom:updated>2025-11-20T10:45:47.882Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IKI8yWU2qRe5HBbJfuEaBA.jpeg" /></figure><p>Europe is setting a new global benchmark for what responsible artificial intelligence looks like.</p><p>Between the EU AI Act, GDPR, Cyber Resilience Act (CRA), and the revised Product Liability Directive (PLD), Europe has set out to make AI systems answer to the people who use them. While the United States treats AI as an engine for innovation and largely trusts the market to self-regulate, and China uses top-down state control to align AI with national priorities, Europe is charting a third path that centres the individual. These regulations mark a fundamental shift away from regulating the technology itself and toward regulating how people interact with AI. For the first time, every actor in the AI ecosystem has both rights and responsibilities.</p><p>The insight driving these regulations is straightforward. AI doesn’t cause harm sitting on a server; it causes harm when people deploy it poorly, design it carelessly, or use it without oversight. Europe’s framework rebalances power by shifting accountability from models to the humans who make those decisions. This isn’t about limiting AI; it’s about making it work for everyone, not just those who build it.</p><p>The implications extend far beyond Europe’s borders. As the world’s second-largest market, the EU’s regulatory framework will inevitably shape global AI design. Compliance with the EU AI Act will become a prerequisite for accessing hundreds of millions of European consumers, meaning these standards will quietly become global defaults; much as how the GDPR redefined data privacy worldwide.</p><p>Here, we break down the real implications of these regulations, what you can expect as a user, what companies are required to deliver, and who must answer when failures happen.</p><h3>Foundation Models: Ending the Trust-Us Era</h3><p>At the core are the foundation model providers: companies like OpenAI, Anthropic, or Google DeepMind, who provide the fundamental capabilities that enable countless applications. For years, their systems have operated as black boxes, but the EU AI Act now recognises these providers as gatekeepers and, for the first time globally, transparency and continuous risk management are now mandatory.</p><p>Foundation model providers now need to publish summaries of their training data, describe capabilities and limitations, and openly disclose known risks. This shifts the paradigm from “trust us, it’s proprietary” to ongoing accountability for what’s been built and how it performs. This transparency also extends beyond one-time disclosure, as providers must then continue to actively identify, test, and mitigate these risks throughout the model’s lifecycle.</p><p>The EU’s existing data protection framework, the GDPR, is then reinforcing this through three essential safeguards: data minimisation, ensuring only necessary data is collected and used; accuracy, requiring training data to be correct and current; and fair and lawful processing, meaning data must be collected transparently, with clear legal basis and without discrimination. This prevents <em>data laundering</em> where developers sweep vast amounts of low-quality or unlawfully obtained data into training sets and claim the model <em>cleans</em> it.</p><p>Meanwhile, the Cyber Resilience Act extends Europe’s long-standing product-safety principles to digital technologies. Software, including foundational models, must now meet security-by-design standards and receive regular updates to address emerging vulnerabilities. The revised Product Liability Directive then closes the loop by treating AI systems as products in their own right, just like cars or medical devices. If an AI model causes harm, the provider can be held liable even without proof of negligence.</p><h3>Application Providers: Accountability in Action</h3><p>Foundation models are abstract. Application providers, developers who turn them into hiring algorithms or medical diagnostics, are the ones who bring AI into everyday life. Europe’s regulations recognise this distinction by imposing requirements that reflect the real-world stakes of deployment.</p><p>While application providers inherit the obligations of foundation model providers, they also face additional requirements when implementing AI in high-risk contexts. AI systems used in employment, credit scoring, or education must now pass conformity assessment before release. Once certified, they carry the CE marking: Europe’s badge of safety, which users can look for as a visible assurance of compliance.</p><p>Beyond certification, providers must maintain audit-ready documentation that records every stage of system development, from training to bias testing and deployment. This creates a traceable chain of accountability, meaning if something goes wrong, the decision can be clearly understood and investigated.</p><h3>Deployers: Keeping Humans in Control</h3><p>Deployers are the organisations that put AI to work in the real world. The EU AI Act assigns them a critical responsibility: use systems only as intended. Deploy a credit-scoring algorithm for hiring decisions, or repurpose a customer service chatbot for resume screening, and you may inherit the same obligations as the original provider, including liability.</p><p>For high-risk AI systems, the AI Act goes further, mandating that decisions remain understandable and open to human intervention. When an algorithm denies someone a loan or rejects a job application a human must be able to examine how that decision was reached, understand its rationale, and, if necessary, override it. This ensures that automation never fully displaces human judgment, particularly in areas that affect people’s rights and opportunities.</p><p>Finally, deployers also share responsibility with the providers to maintain detailed logs and notify regulators if an AI system malfunctions or causes harm.</p><h3>End Users: Empowered to Challenge</h3><p>For too long, people have been ranked, filtered, and judged by algorithms they couldn’t see, question, or challenge. The EU’s framework changes this dynamic by making transparency and contestability legal requirements. In high-risk situations, the GDPR already grants the right to object to purely automated decision-making, and the AI Act reinforces this by requiring deployers to enable users to challenge AI-based outcomes. Denied a job or loan because of an algorithm? You can now demand an explanation and contest the decision.</p><p>Beyond this, the framework also establishes concrete accountability mechanisms. The Product Liability Directive ensures individuals can seek compensation for harm caused by AI systems, including through faulty software updates. This closes long-standing gaps where algorithmic errors previously fell into legal grey zones.</p><p>These regulations work together as an integrated system: transparency enables challenge, challenge enables remedy, and security ensures the whole framework rests on trustworthy foundations.</p><h3>Building Trustworthy AI at Scale</h3><p>Europe’s approach focuses regulation on how AI is used, governed, and experienced, laying the foundation for a sustainable AI ecosystem where compliance is not a constraint, but a badge of trust. As global companies adapt their systems to meet European standards, this model of technological responsibility, one that prioritizes individual rights over market velocity or state control, will increasingly define how AI operates worldwide.</p><p>The Product Liability Directive ensures justice for people harmed by AI. The Cyber Resilience Act makes security a shared responsibility. The GDPR anchors privacy as a human right. And the AI Act ties it all together, making fairness, transparency, and oversight foundational to how AI is built and deployed across the entire ecosystem.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6722e48853ba" width="1" height="1" alt=""><hr><p><a href="https://medium.com/stupid-human/competitive-advantage-through-regulation-europes-vision-for-responsible-ai-6722e48853ba">Competitive Advantage Through Regulation: Europe’s Vision for Responsible AI</a> was originally published in <a href="https://medium.com/stupid-human">Stupid Human</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>