<?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:itunes="http://www.itunes.com/dtds/podcast-1.0.dtd" xmlns:googleplay="http://www.google.com/schemas/play-podcasts/1.0"><channel><title><![CDATA[makeyourmind/up: mind.the.gap]]></title><description><![CDATA[What's coming in AI, before it's obvious.]]></description><link>https://home.makeyourmindup.ai/s/mindthegap</link><image><url>https://substackcdn.com/image/fetch/$s_!paPg!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F72aee8ff-74d4-4414-824e-b09670cc1618_600x600.png</url><title>makeyourmind/up: mind.the.gap</title><link>https://home.makeyourmindup.ai/s/mindthegap</link></image><generator>Substack</generator><lastBuildDate>Sat, 21 Nov 2026 08:52:15 GMT</lastBuildDate><atom:link href="https://home.makeyourmindup.ai/feed" rel="self" type="application/rss+xml"/><copyright><![CDATA[makeyourmindup]]></copyright><language><![CDATA[en]]></language><webMaster><![CDATA[mindmake@substack.com]]></webMaster><itunes:owner><itunes:email><![CDATA[mindmake@substack.com]]></itunes:email><itunes:name><![CDATA[Krish]]></itunes:name></itunes:owner><itunes:author><![CDATA[Krish]]></itunes:author><googleplay:owner><![CDATA[mindmake@substack.com]]></googleplay:owner><googleplay:email><![CDATA[mindmake@substack.com]]></googleplay:email><googleplay:author><![CDATA[Krish]]></googleplay:author><itunes:block><![CDATA[Yes]]></itunes:block><item><title><![CDATA[Meta staff were ranked by AI use. They gamed it.]]></title><description><![CDATA[Watch now (5 mins) | Big tech companies started ranking staff by how much AI they used, and staff set AI to work on nothing to climb the charts. Why counting AI use backfires, and what to count instead.]]></description><link>https://home.makeyourmindup.ai/p/meta-staff-were-ranked-by-ai-use</link><guid isPermaLink="false">https://home.makeyourmindup.ai/p/meta-staff-were-ranked-by-ai-use</guid><dc:creator><![CDATA[Krish]]></dc:creator><pubDate>Fri, 13 Nov 2026 17:31:02 GMT</pubDate><enclosure url="https://api.substack.com/feed/podcast/224515085/3912e44fef15ed6155d487baf1359f76.mp3" length="0" type="audio/mpeg"/><content:encoded><![CDATA[<p>At some of the biggest companies in tech, staff were judged, and then ranked, by how much AI they used. The staff learned fast. They used more AI. A lot of it on nothing.</p><p>We wanted to know why that happens, and what to count instead. The reason turns out to be more than a century old, and it starts with rats.</p><h2>A CENT A RAT TAIL</h2><p>In 1902, Hanoi, then run by France, was afraid of plague, and plague travels with rats. So the French set a bounty: one cent per rat, according to <a href="https://www.atlasobscura.com/articles/hanoi-rat-massacre-1902">the history and travel site Atlas Obscura</a>, drawing on historian Michael Vann&#8217;s research. To claim it, you handed in the rat&#8217;s tail, which spared the town hall a pile of bodies.</p><p>Tails poured in. Then officials spotted something odd: tailless rats, scampering about, very much alive. Hanoi&#8217;s rat catchers had worked it out. Snip the tail, let the rat go, and it breeds you more tails. Then health inspectors found rat farms in the countryside just outside Hanoi.</p><p>Hanoi paid for tails, and tails is what it got, plus a side business in rats.</p><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!85t6!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1183e86e-9699-4cdb-8396-04c789670e5c_1360x762.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!85t6!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1183e86e-9699-4cdb-8396-04c789670e5c_1360x762.png 424w, https://substackcdn.com/image/fetch/$s_!85t6!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1183e86e-9699-4cdb-8396-04c789670e5c_1360x762.png 848w, https://substackcdn.com/image/fetch/$s_!85t6!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1183e86e-9699-4cdb-8396-04c789670e5c_1360x762.png 1272w, https://substackcdn.com/image/fetch/$s_!85t6!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1183e86e-9699-4cdb-8396-04c789670e5c_1360x762.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!85t6!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1183e86e-9699-4cdb-8396-04c789670e5c_1360x762.png" width="1360" height="762" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/1183e86e-9699-4cdb-8396-04c789670e5c_1360x762.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:762,&quot;width&quot;:1360,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;How a rat bounty made more rats. Hanoi, 1902: the French paid one cent per rat, with the tail as proof. People snipped the tails and let the rats go, and health inspectors found rat farms outside Hanoi. The city wanted fewer rats; it counted tails.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="How a rat bounty made more rats. Hanoi, 1902: the French paid one cent per rat, with the tail as proof. People snipped the tails and let the rats go, and health inspectors found rat farms outside Hanoi. The city wanted fewer rats; it counted tails." title="How a rat bounty made more rats. Hanoi, 1902: the French paid one cent per rat, with the tail as proof. People snipped the tails and let the rats go, and health inspectors found rat farms outside Hanoi. The city wanted fewer rats; it counted tails." srcset="https://substackcdn.com/image/fetch/$s_!85t6!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1183e86e-9699-4cdb-8396-04c789670e5c_1360x762.png 424w, https://substackcdn.com/image/fetch/$s_!85t6!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1183e86e-9699-4cdb-8396-04c789670e5c_1360x762.png 848w, https://substackcdn.com/image/fetch/$s_!85t6!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1183e86e-9699-4cdb-8396-04c789670e5c_1360x762.png 1272w, https://substackcdn.com/image/fetch/$s_!85t6!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1183e86e-9699-4cdb-8396-04c789670e5c_1360x762.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><p>There&#8217;s the whole mistake, in miniature. Pay people for a sign of the thing you want, and they&#8217;ll get very good at making signs. Economists call it Goodhart&#8217;s law: once a number becomes a target, it stops telling you much. We call it the rat-tail problem.</p><h2>THE MANDATES</h2><p>With AI, it began with bosses judging staff on their AI use.</p><p>In April 2025, Shopify&#8217;s chief executive, Tobi L&#252;tke, wrote to staff that &#8220;<a href="https://x.com/tobi/status/1909251946235437514">reflexive AI usage is now a baseline expectation</a>&#8220;, and that AI use would be part of performance reviews. Later that month, Luis von Ahn, who runs the language app Duolingo, told staff that AI use would be weighed in their reviews too. <a href="https://www.efinancialcareers.com/news/coinbase-fired-an-engineer-after-they-refused-to-use-ai-coding-tools">Coinbase&#8217;s boss, Brian Armstrong</a>, said an engineer who refused to use the company&#8217;s AI coding tools was fired, eFinancialCareers reported in March 2026. And at Meta, counts of AI use went into staff performance reviews, <a href="https://daily.dev/posts/meta-drops-ai-token-usage-from-engineer-performance-reviews-after-employees-gamed-the-metric-lcwyhjcee">according to daily.dev</a>, which later reported that Meta had dropped them.</p><p>We think these bosses were right that AI matters. Our quarrel is with one small word: &#8220;use&#8221;. Judge people on AI use and somebody has to measure it. And the cheapest thing to measure is simply how much got used.</p><h2>THE LEADERBOARDS</h2><p>AI use is measured in tokens, which Fortune describes as &#8220;the basic units of data or words that AI models process&#8221;. Every one of them costs money, so a token count is really a bill.</p><p>At Meta, the owner of Facebook, an employee built an internal leaderboard that tracked the tokens used by the company&#8217;s more than 85,000 staff, <a href="https://fortune.com/2026/04/09/meta-killed-employee-ai-token-dashboard/">as Fortune reported in April 2026</a>. It ranked the top 250 users and gave them titles. Real ones. &#8220;Token Legend&#8221;. &#8220;Cache Wizard&#8221;. Over 30 days, usage on the dashboard passed 60 trillion tokens. By Fortune&#8217;s sums, at $5 for every million tokens, the top user alone could have cost Meta more than $1.4 million. Two days after news of the board broke, the whole thing went dark.</p><p>Amazon had one too. It was called KiroRank, after Kiro, Amazon&#8217;s AI tool for writing code. <a href="https://www.cio.com/article/4178825/amazon-deletes-devs-tokenmaxxing-leaderboard-to-minimize-costs-2.html">According to the tech news site CIO</a>, some staff set AI agents (helpers that keep working on a task by themselves, without a person typing) to do unnecessary tasks, just to push their scores up. It was taken offline, and a senior vice president said the computing bills it ran up were too high.</p><p>That&#8217;s rat farming with a laptop: nobody needed the tasks, only the tokens. And it&#8217;s the perfect way to climb a chart that counts usage.</p><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!FgXJ!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff9a32c34-7405-47f6-92e4-93fcb52b85fb_1360x904.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!FgXJ!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff9a32c34-7405-47f6-92e4-93fcb52b85fb_1360x904.png 424w, https://substackcdn.com/image/fetch/$s_!FgXJ!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff9a32c34-7405-47f6-92e4-93fcb52b85fb_1360x904.png 848w, https://substackcdn.com/image/fetch/$s_!FgXJ!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff9a32c34-7405-47f6-92e4-93fcb52b85fb_1360x904.png 1272w, https://substackcdn.com/image/fetch/$s_!FgXJ!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff9a32c34-7405-47f6-92e4-93fcb52b85fb_1360x904.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!FgXJ!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff9a32c34-7405-47f6-92e4-93fcb52b85fb_1360x904.png" width="1360" height="904" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/f9a32c34-7405-47f6-92e4-93fcb52b85fb_1360x904.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:904,&quot;width&quot;:1360,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;The token leaderboards. A Meta employee's board tracked the tokens used by more than 85,000 Meta staff and ranked the top 250, with titles such as Token Legend and Cache Wizard. Over 30 days usage passed 60 trillion tokens; the top user alone could have cost Meta more than $1.4 million, by Fortune's sums. Meta's board went down two days after the news broke; Amazon's KiroRank, where staff set AI agents on unnecessary tasks, was taken offline over its computing bills.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="The token leaderboards. A Meta employee's board tracked the tokens used by more than 85,000 Meta staff and ranked the top 250, with titles such as Token Legend and Cache Wizard. Over 30 days usage passed 60 trillion tokens; the top user alone could have cost Meta more than $1.4 million, by Fortune's sums. Meta's board went down two days after the news broke; Amazon's KiroRank, where staff set AI agents on unnecessary tasks, was taken offline over its computing bills." title="The token leaderboards. A Meta employee's board tracked the tokens used by more than 85,000 Meta staff and ranked the top 250, with titles such as Token Legend and Cache Wizard. Over 30 days usage passed 60 trillion tokens; the top user alone could have cost Meta more than $1.4 million, by Fortune's sums. Meta's board went down two days after the news broke; Amazon's KiroRank, where staff set AI agents on unnecessary tasks, was taken offline over its computing bills." srcset="https://substackcdn.com/image/fetch/$s_!FgXJ!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff9a32c34-7405-47f6-92e4-93fcb52b85fb_1360x904.png 424w, https://substackcdn.com/image/fetch/$s_!FgXJ!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff9a32c34-7405-47f6-92e4-93fcb52b85fb_1360x904.png 848w, https://substackcdn.com/image/fetch/$s_!FgXJ!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff9a32c34-7405-47f6-92e4-93fcb52b85fb_1360x904.png 1272w, https://substackcdn.com/image/fetch/$s_!FgXJ!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff9a32c34-7405-47f6-92e4-93fcb52b85fb_1360x904.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><h2>THE U-TURNS</h2><p>Then three companies backed down: Duolingo, Meta and Amazon.</p><p><a href="https://www.wpxi.com/news/local/duolingo-ceo-says-company-backtracked-ai-evaluations-performance-reviews/6PP4GFMW4ZC6LLAKUMAOQELQSE/">Duolingo&#8217;s von Ahn told a podcast in April 2026</a> that AI use was no longer part of performance reviews, the Pittsburgh news site WPXI reported. &#8220;For a while, it was part of performance reviews, but we decided not to do that,&#8221; he said. In September 2026, Meta said it would no longer use token counts or AI adoption dashboards to judge performance, according to daily.dev, a news site for software developers. In late October, <a href="https://www.cnbc.com/2026/10/27/amazon-ai-coding-tokenmaxxing.html">CNBC reported</a> that Amazon had told its engineers that token counts and similar numbers are &#8220;deeply flawed&#8221; ways to measure teams&#8217; AI use. The memo also said: &#8220;If your work does not benefit from current AI tooling, that is a finding, not a failing.&#8221;</p><p>To us, that&#8217;s the most important line in the whole story. Under a usage target, saying &#8220;AI didn&#8217;t help here&#8221; sounds like admitting you&#8217;re behind. So people don&#8217;t say it. They keep the chart looking healthy, and the company loses the one piece of information it actually needs: where AI works and where it doesn&#8217;t.</p><p>Zoom out, and here&#8217;s what we think it means if you manage people. A usage number tells you how much your company is paying for AI. It tells you nothing about what your company is getting for it. If you judge people on the bill, they will run up the bill. And you will have no idea whether the work got any better.</p><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!7tgC!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F995b8d62-5914-4aac-a255-16b458bc9319_1360x732.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!7tgC!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F995b8d62-5914-4aac-a255-16b458bc9319_1360x732.png 424w, https://substackcdn.com/image/fetch/$s_!7tgC!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F995b8d62-5914-4aac-a255-16b458bc9319_1360x732.png 848w, https://substackcdn.com/image/fetch/$s_!7tgC!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F995b8d62-5914-4aac-a255-16b458bc9319_1360x732.png 1272w, https://substackcdn.com/image/fetch/$s_!7tgC!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F995b8d62-5914-4aac-a255-16b458bc9319_1360x732.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!7tgC!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F995b8d62-5914-4aac-a255-16b458bc9319_1360x732.png" width="1360" height="732" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/995b8d62-5914-4aac-a255-16b458bc9319_1360x732.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:732,&quot;width&quot;:1360,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;The U-turns. Duolingo, April 2026: its boss said AI use was no longer part of performance reviews. Meta, September 2026: said it would stop using token counts and AI adoption dashboards to judge performance. Amazon, October 2026: told engineers token counts are deeply flawed ways to measure teams' AI use, and that AI not helping is a finding, not a failing.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="The U-turns. Duolingo, April 2026: its boss said AI use was no longer part of performance reviews. Meta, September 2026: said it would stop using token counts and AI adoption dashboards to judge performance. Amazon, October 2026: told engineers token counts are deeply flawed ways to measure teams' AI use, and that AI not helping is a finding, not a failing." title="The U-turns. Duolingo, April 2026: its boss said AI use was no longer part of performance reviews. Meta, September 2026: said it would stop using token counts and AI adoption dashboards to judge performance. Amazon, October 2026: told engineers token counts are deeply flawed ways to measure teams' AI use, and that AI not helping is a finding, not a failing." srcset="https://substackcdn.com/image/fetch/$s_!7tgC!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F995b8d62-5914-4aac-a255-16b458bc9319_1360x732.png 424w, https://substackcdn.com/image/fetch/$s_!7tgC!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F995b8d62-5914-4aac-a255-16b458bc9319_1360x732.png 848w, https://substackcdn.com/image/fetch/$s_!7tgC!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F995b8d62-5914-4aac-a255-16b458bc9319_1360x732.png 1272w, https://substackcdn.com/image/fetch/$s_!7tgC!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F995b8d62-5914-4aac-a255-16b458bc9319_1360x732.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><h2>HALF THE TARGET, MORE SALES</h2><p>Our favourite evidence comes from a study of a company that kept counting, and simply asked for less.</p><p><a href="https://arxiv.org/abs/2609.32859v1">A study posted to arXiv</a> (a site where researchers share papers) in September 2026 looked at a large medical-device company that told roughly 5,000 of its employees to submit at least 200 queries a month to an in-house AI assistant. Query counts bunched right at the target, a sign people asked just enough to hit it. And 31% of the queries were repeated or off-task.</p><p>Then the firm lowered the target to 100 queries a month. It did this one branch at a time, which let the researchers compare branches on the new target with branches still on the old one. Total queries fell by about 30%. (Two numbers here look alike: 31% was the share of junk under the old target, and 30% is how much the total fell once the target was halved.) About 90% of that drop was the junk: the repeats and the off-topic queries. The useful queries mostly stayed.</p><p>Among sales staff, monthly sales went up by 7%, the researchers estimate, comparing branches on the new target with those still on the old one. The paper is a preprint, so other researchers haven&#8217;t checked it yet.</p><p>Why? Our guess is simple. People got back the time they had been spending feeding the target, and spent it selling.</p><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!_viH!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Faa6ed3a0-dfd8-421e-9037-c9b9a3ac3f06_1360x662.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!_viH!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Faa6ed3a0-dfd8-421e-9037-c9b9a3ac3f06_1360x662.png 424w, https://substackcdn.com/image/fetch/$s_!_viH!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Faa6ed3a0-dfd8-421e-9037-c9b9a3ac3f06_1360x662.png 848w, https://substackcdn.com/image/fetch/$s_!_viH!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Faa6ed3a0-dfd8-421e-9037-c9b9a3ac3f06_1360x662.png 1272w, https://substackcdn.com/image/fetch/$s_!_viH!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Faa6ed3a0-dfd8-421e-9037-c9b9a3ac3f06_1360x662.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!_viH!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Faa6ed3a0-dfd8-421e-9037-c9b9a3ac3f06_1360x662.png" width="1360" height="662" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/aa6ed3a0-dfd8-421e-9037-c9b9a3ac3f06_1360x662.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:662,&quot;width&quot;:1360,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;Half the target, more sales. A medical-device company told about 5,000 staff to send its AI assistant at least 200 queries a month; 31% were repeated or off-task. It cut the target to 100, branch by branch. Total queries fell 30%, about 90% of the drop junk, and monthly sales among sales staff rose an estimated 7%.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="Half the target, more sales. A medical-device company told about 5,000 staff to send its AI assistant at least 200 queries a month; 31% were repeated or off-task. It cut the target to 100, branch by branch. Total queries fell 30%, about 90% of the drop junk, and monthly sales among sales staff rose an estimated 7%." title="Half the target, more sales. A medical-device company told about 5,000 staff to send its AI assistant at least 200 queries a month; 31% were repeated or off-task. It cut the target to 100, branch by branch. Total queries fell 30%, about 90% of the drop junk, and monthly sales among sales staff rose an estimated 7%." srcset="https://substackcdn.com/image/fetch/$s_!_viH!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Faa6ed3a0-dfd8-421e-9037-c9b9a3ac3f06_1360x662.png 424w, https://substackcdn.com/image/fetch/$s_!_viH!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Faa6ed3a0-dfd8-421e-9037-c9b9a3ac3f06_1360x662.png 848w, https://substackcdn.com/image/fetch/$s_!_viH!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Faa6ed3a0-dfd8-421e-9037-c9b9a3ac3f06_1360x662.png 1272w, https://substackcdn.com/image/fetch/$s_!_viH!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Faa6ed3a0-dfd8-421e-9037-c9b9a3ac3f06_1360x662.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><h2>THE COBRA THAT NEVER WAS</h2><p>You might know a snakier version of this tale, about a British bounty on cobras in Delhi.</p><p>We love that story. <a href="https://www.thejuggernaut.com/cobra-effect-history-british-raj-india">The Juggernaut, a publication about South Asia</a>, calls it &#8220;the made-up story of a British incentive to curb Delhi&#8217;s snake population&#8221;, with &#8220;little evidence to support the veracity&#8221; of it. The phrase &#8220;cobra effect&#8221; itself comes from a German economist, Horst Siebert, in a book from 2001.</p><p>So even the most famous story about counting the wrong thing got repeated for years because it sounded right, with nobody checking. Which is a fair warning for anyone about to build a dashboard. A number on a screen feels like proof. It&#8217;s only as true as what people did to produce it.</p><p>Hanoi counted tails and got tails. At Meta and Amazon, staff ran unnecessary tasks just to climb token leaderboards. The medical-device company asked for fewer queries, and its sales staff sold 7% more. Whatever you count, people learn to produce. So count what got done.</p><p>If you&#8217;re counting how much AI your people use, you&#8217;re counting rat tails.</p><h2>3 things you can do tomorrow</h2><ul><li><p>Scrap any target or ranking for how much AI people use. If it&#8217;s in a review form, a dashboard or a team chart, take it out.</p></li><li><p>Make &#8220;AI didn&#8217;t help here&#8221; a good answer. Amazon now calls it &#8220;a finding, not a failing&#8221;. Say the same to your team, so nobody feels they have to fake it.</p></li><li><p>Judge AI on a number you already track, like sales, tickets closed or invoices sent. If that number hasn&#8217;t moved, the AI isn&#8217;t working yet, whatever the usage says.</p></li></ul><h2>SOURCES</h2><ol><li><p><a href="https://www.atlasobscura.com/articles/hanoi-rat-massacre-1902">Atlas Obscura: The Great Hanoi Rat Massacre of 1902</a></p></li><li><p><a href="https://x.com/tobi/status/1909251946235437514">Tobi L&#252;tke on X: Reflexive AI usage is now a baseline expectation at Shopify (7 Apr 2025)</a></p></li><li><p><a href="https://www.efinancialcareers.com/news/coinbase-fired-an-engineer-after-they-refused-to-use-ai-coding-tools">eFinancialCareers: Coinbase fired an engineer after they refused to use AI coding tools (Mar 2026)</a></p></li><li><p><a href="https://daily.dev/posts/meta-drops-ai-token-usage-from-engineer-performance-reviews-after-employees-gamed-the-metric-lcwyhjcee">daily.dev: Meta drops AI token usage from engineer performance reviews (Sep 2026)</a></p></li><li><p><a href="https://fortune.com/2026/04/09/meta-killed-employee-ai-token-dashboard/">Fortune: A Meta employee created a dashboard so coworkers can compete to be the company&#8217;s No. 1 AI token user (9 Apr 2026)</a></p></li><li><p><a href="https://www.cio.com/article/4178825/amazon-deletes-devs-tokenmaxxing-leaderboard-to-minimize-costs-2.html">CIO: Amazon deletes devs&#8217; tokenmaxxing leaderboard to minimize costs (29 May 2026)</a></p></li><li><p><a href="https://www.wpxi.com/news/local/duolingo-ceo-says-company-backtracked-ai-evaluations-performance-reviews/6PP4GFMW4ZC6LLAKUMAOQELQSE/">WPXI: Duolingo CEO says company backtracked on AI evaluations in performance reviews (Apr 2026)</a></p></li><li><p><a href="https://www.cnbc.com/2026/10/27/amazon-ai-coding-tokenmaxxing.html">CNBC: Amazon&#8217;s anti-tokenmaxxing memo (27 Oct 2026)</a></p></li><li><p><a href="https://arxiv.org/abs/2609.32859v1">arXiv 2609.32859: AI usage targets at a medical-device company (Sep 2026)</a></p></li><li><p><a href="https://www.thejuggernaut.com/cobra-effect-history-british-raj-india">The Juggernaut: The Cobra Effect and the Colonial Gaze</a></p></li></ol><p><em>Checked on 11 November 2026 against the sources below.</em></p>]]></content:encoded></item><item><title><![CDATA[Why every country wants its own AI, and what that means for your business]]></title><description><![CDATA[Watch now (5 mins) | Every AI has two passports: where it was built and where it runs. We tested three on work questions to see what each one changes.]]></description><link>https://home.makeyourmindup.ai/p/why-every-country-wants-its-own-ai</link><guid isPermaLink="false">https://home.makeyourmindup.ai/p/why-every-country-wants-its-own-ai</guid><pubDate>Fri, 06 Nov 2026 17:31:13 GMT</pubDate><enclosure url="https://api.substack.com/feed/podcast/223757154/a360b19eb4c08209cd825f9c38ced260.mp3" length="0" type="audio/mpeg"/><content:encoded><![CDATA[<p>Every country wants its own AI now, and the reason sits inside the AI your team already uses.</p><p>Most of us pick an AI because it works and the price is fine. But Airbnb&#8217;s boss, a popular coding tool and a fast-growing American startup have all said, in their own words, that they rely on AI built in China. Meanwhile Britain, France, the Gulf, India and Korea are spending serious money so they don&#8217;t have to rely on anyone else&#8217;s.</p><p>So we spent the week working out whether it matters which country your AI comes from.</p><h2>EVERY COUNTRY WANTS ITS OWN AI</h2><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!Kesd!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffb3423ee-d98a-4ee0-9430-7cb067aafe6c_2400x850.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!Kesd!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffb3423ee-d98a-4ee0-9430-7cb067aafe6c_2400x850.png 424w, https://substackcdn.com/image/fetch/$s_!Kesd!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffb3423ee-d98a-4ee0-9430-7cb067aafe6c_2400x850.png 848w, https://substackcdn.com/image/fetch/$s_!Kesd!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffb3423ee-d98a-4ee0-9430-7cb067aafe6c_2400x850.png 1272w, https://substackcdn.com/image/fetch/$s_!Kesd!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffb3423ee-d98a-4ee0-9430-7cb067aafe6c_2400x850.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!Kesd!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffb3423ee-d98a-4ee0-9430-7cb067aafe6c_2400x850.png" width="1456" height="516" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/fb3423ee-d98a-4ee0-9430-7cb067aafe6c_2400x850.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:516,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;Quote from Stanford's AI Index 2026: the US produced 59 notable AI models in 2025 to China's 35, while China leads on research papers and patents.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="Quote from Stanford's AI Index 2026: the US produced 59 notable AI models in 2025 to China's 35, while China leads on research papers and patents." title="Quote from Stanford's AI Index 2026: the US produced 59 notable AI models in 2025 to China's 35, while China leads on research papers and patents." srcset="https://substackcdn.com/image/fetch/$s_!Kesd!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffb3423ee-d98a-4ee0-9430-7cb067aafe6c_2400x850.png 424w, https://substackcdn.com/image/fetch/$s_!Kesd!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffb3423ee-d98a-4ee0-9430-7cb067aafe6c_2400x850.png 848w, https://substackcdn.com/image/fetch/$s_!Kesd!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffb3423ee-d98a-4ee0-9430-7cb067aafe6c_2400x850.png 1272w, https://substackcdn.com/image/fetch/$s_!Kesd!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffb3423ee-d98a-4ee0-9430-7cb067aafe6c_2400x850.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><p>Start with who&#8217;s in the race. Stanford counts the most important new AI models each year, and in 2025 America made 59 and China made 35. At the very top it&#8217;s closer than that sounds: by March 2026, Stanford found, Anthropic&#8217;s best model led the best Chinese one by just 2.7%.</p><p>Sovereign AI is the name for a country making sure it can rely on an AI of its own.</p><p>Everyone else wants a seat. The UK has a government-backed Sovereign AI fund. France has Mistral. In the Gulf, the UAE&#8217;s MBZUAI, an AI university, and G42, an AI company, launched K2 Think, and Saudi Arabia&#8217;s HUMAIN launched an Arabic chat app. India picked a company called Sarvam to build its national model. South Korea runs a national AI project of its own.</p><p>It&#8217;s harder than it looks. South Korea dropped Naver Cloud from that national project because a piece of its model turned out to be borrowed from Alibaba&#8217;s Qwen, a Chinese model. A national AI with an imported part, and a fair preview of the problem for everyone else.</p><h2>THE AI YOU USE ALREADY CROSSES BORDERS</h2><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!0KAD!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F23a816e6-84c9-486c-9011-c7e963b2d8a6_2400x766.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!0KAD!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F23a816e6-84c9-486c-9011-c7e963b2d8a6_2400x766.png 424w, https://substackcdn.com/image/fetch/$s_!0KAD!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F23a816e6-84c9-486c-9011-c7e963b2d8a6_2400x766.png 848w, https://substackcdn.com/image/fetch/$s_!0KAD!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F23a816e6-84c9-486c-9011-c7e963b2d8a6_2400x766.png 1272w, https://substackcdn.com/image/fetch/$s_!0KAD!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F23a816e6-84c9-486c-9011-c7e963b2d8a6_2400x766.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!0KAD!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F23a816e6-84c9-486c-9011-c7e963b2d8a6_2400x766.png" width="1456" height="465" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/23a816e6-84c9-486c-9011-c7e963b2d8a6_2400x766.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:465,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;Quote from Fortune, 21 October 2025: Airbnb's boss Brian Chesky says the company relies a lot on Alibaba's Qwen model because it is good, fast and cheap.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="Quote from Fortune, 21 October 2025: Airbnb's boss Brian Chesky says the company relies a lot on Alibaba's Qwen model because it is good, fast and cheap." title="Quote from Fortune, 21 October 2025: Airbnb's boss Brian Chesky says the company relies a lot on Alibaba's Qwen model because it is good, fast and cheap." srcset="https://substackcdn.com/image/fetch/$s_!0KAD!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F23a816e6-84c9-486c-9011-c7e963b2d8a6_2400x766.png 424w, https://substackcdn.com/image/fetch/$s_!0KAD!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F23a816e6-84c9-486c-9011-c7e963b2d8a6_2400x766.png 848w, https://substackcdn.com/image/fetch/$s_!0KAD!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F23a816e6-84c9-486c-9011-c7e963b2d8a6_2400x766.png 1272w, https://substackcdn.com/image/fetch/$s_!0KAD!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F23a816e6-84c9-486c-9011-c7e963b2d8a6_2400x766.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><p>Businesses are making the same choice countries are, mostly on price.</p><p>Airbnb&#8217;s boss, Brian Chesky, told Bloomberg last October, as Fortune reported, that its customer service leans on Alibaba&#8217;s Qwen: &#8220;We&#8217;re relying a lot on Alibaba&#8217;s Qwen model. It&#8217;s very good. It&#8217;s also fast and cheap.&#8221;</p><p>Cursor, a coding tool lots of software teams use, built its newest model on top of Kimi, from China&#8217;s Moonshot AI, and didn&#8217;t say so at first. Once people spotted it, co-founder Aman Sanger said: &#8220;It was a miss to not mention the Kimi base in our blog from the start.&#8221;</p><p>And Lindy, an American startup, told NPR it had moved all its traffic onto DeepSeek. Its boss told NPR its biggest bill had been Anthropic, and DeepSeek &#8220;was just 10x cheaper.&#8221;</p><p>It runs the other way too. HSBC, a British bank, signed a deal with France&#8217;s Mistral. Everyone in this story is doing the ordinary thing: shopping.</p><h2>WHERE IT WAS BUILT SHAPES WHAT IT SAYS</h2><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!Fap_!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F53ffa065-c4f8-43a9-8064-463567cf30e8_2400x1060.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!Fap_!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F53ffa065-c4f8-43a9-8064-463567cf30e8_2400x1060.png 424w, https://substackcdn.com/image/fetch/$s_!Fap_!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F53ffa065-c4f8-43a9-8064-463567cf30e8_2400x1060.png 848w, https://substackcdn.com/image/fetch/$s_!Fap_!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F53ffa065-c4f8-43a9-8064-463567cf30e8_2400x1060.png 1272w, https://substackcdn.com/image/fetch/$s_!Fap_!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F53ffa065-c4f8-43a9-8064-463567cf30e8_2400x1060.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!Fap_!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F53ffa065-c4f8-43a9-8064-463567cf30e8_2400x1060.png" width="1456" height="643" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/53ffa065-c4f8-43a9-8064-463567cf30e8_2400x1060.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:643,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;Quote from the Christian Science Monitor, 23 September 2026: on 200 news questions, Chinese-owned chatbots refused, used government language or got it wrong in 35% of cases, 63% for DeepSeek, while Claude stayed factual.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="Quote from the Christian Science Monitor, 23 September 2026: on 200 news questions, Chinese-owned chatbots refused, used government language or got it wrong in 35% of cases, 63% for DeepSeek, while Claude stayed factual." title="Quote from the Christian Science Monitor, 23 September 2026: on 200 news questions, Chinese-owned chatbots refused, used government language or got it wrong in 35% of cases, 63% for DeepSeek, while Claude stayed factual." srcset="https://substackcdn.com/image/fetch/$s_!Fap_!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F53ffa065-c4f8-43a9-8064-463567cf30e8_2400x1060.png 424w, https://substackcdn.com/image/fetch/$s_!Fap_!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F53ffa065-c4f8-43a9-8064-463567cf30e8_2400x1060.png 848w, https://substackcdn.com/image/fetch/$s_!Fap_!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F53ffa065-c4f8-43a9-8064-463567cf30e8_2400x1060.png 1272w, https://substackcdn.com/image/fetch/$s_!Fap_!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F53ffa065-c4f8-43a9-8064-463567cf30e8_2400x1060.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!ApaR!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd97ae828-fd50-4163-90b9-44a25da0eed2_2400x1370.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!ApaR!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd97ae828-fd50-4163-90b9-44a25da0eed2_2400x1370.png 424w, https://substackcdn.com/image/fetch/$s_!ApaR!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd97ae828-fd50-4163-90b9-44a25da0eed2_2400x1370.png 848w, https://substackcdn.com/image/fetch/$s_!ApaR!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd97ae828-fd50-4163-90b9-44a25da0eed2_2400x1370.png 1272w, https://substackcdn.com/image/fetch/$s_!ApaR!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd97ae828-fd50-4163-90b9-44a25da0eed2_2400x1370.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!ApaR!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd97ae828-fd50-4163-90b9-44a25da0eed2_2400x1370.png" width="1456" height="831" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/d97ae828-fd50-4163-90b9-44a25da0eed2_2400x1370.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:831,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;Every AI has two passports. Passport one, where it was born, shapes what it tells you. Passport two, whose computers it runs on, decides who sees what you type. DeepSeek's app was born in China and lives in China. DeepSeek downloaded and run at home lives on your own computers, but its answers came along. Mistral at HSBC was born in France and lives on HSBC's own systems.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="Every AI has two passports. Passport one, where it was born, shapes what it tells you. Passport two, whose computers it runs on, decides who sees what you type. DeepSeek's app was born in China and lives in China. DeepSeek downloaded and run at home lives on your own computers, but its answers came along. Mistral at HSBC was born in France and lives on HSBC's own systems." title="Every AI has two passports. Passport one, where it was born, shapes what it tells you. Passport two, whose computers it runs on, decides who sees what you type. DeepSeek's app was born in China and lives in China. DeepSeek downloaded and run at home lives on your own computers, but its answers came along. Mistral at HSBC was born in France and lives on HSBC's own systems." srcset="https://substackcdn.com/image/fetch/$s_!ApaR!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd97ae828-fd50-4163-90b9-44a25da0eed2_2400x1370.png 424w, https://substackcdn.com/image/fetch/$s_!ApaR!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd97ae828-fd50-4163-90b9-44a25da0eed2_2400x1370.png 848w, https://substackcdn.com/image/fetch/$s_!ApaR!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd97ae828-fd50-4163-90b9-44a25da0eed2_2400x1370.png 1272w, https://substackcdn.com/image/fetch/$s_!ApaR!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd97ae828-fd50-4163-90b9-44a25da0eed2_2400x1370.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><p>So here&#8217;s the way we think about it. Every AI has two passports. One says where it was born. The other says where it lives, meaning whose computers it runs on.</p><p>The first passport shapes what it tells you. The Christian Science Monitor put 200 news questions to four Chinese chatbots (DeepSeek, Kimi, Qwen and GLM) and compared them with Claude. In more than a third of cases (35%) the Chinese bots refused, used the Chinese government&#8217;s preferred language, or got it wrong. For DeepSeek it was 63%. Claude, the paper found, &#8220;answered all of them.&#8221;</p><p>America has its own version. Last summer, the developer Simon Willison caught Elon Musk&#8217;s Grok searching X for Elon Musk&#8217;s posts before it answered a question about Israel and Palestine. xAI, Musk&#8217;s AI company, later told it to stop.</p><p>The lesson applies everywhere. Every maker&#8217;s choices travel with its AI, and a model&#8217;s birthplace is one of those choices.</p><h2>WHERE IT RUNS DECIDES WHO SEES YOUR DATA</h2><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!U0Qv!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F54f7151f-536d-48da-85b3-4e34e8481d57_2400x760.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!U0Qv!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F54f7151f-536d-48da-85b3-4e34e8481d57_2400x760.png 424w, https://substackcdn.com/image/fetch/$s_!U0Qv!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F54f7151f-536d-48da-85b3-4e34e8481d57_2400x760.png 848w, https://substackcdn.com/image/fetch/$s_!U0Qv!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F54f7151f-536d-48da-85b3-4e34e8481d57_2400x760.png 1272w, https://substackcdn.com/image/fetch/$s_!U0Qv!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F54f7151f-536d-48da-85b3-4e34e8481d57_2400x760.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!U0Qv!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F54f7151f-536d-48da-85b3-4e34e8481d57_2400x760.png" width="1456" height="461" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/54f7151f-536d-48da-85b3-4e34e8481d57_2400x760.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:461,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;Quote from DeepSeek's privacy policy, read 30 October 2026: it collects, processes and stores your personal data in the People's Republic of China.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="Quote from DeepSeek's privacy policy, read 30 October 2026: it collects, processes and stores your personal data in the People's Republic of China." title="Quote from DeepSeek's privacy policy, read 30 October 2026: it collects, processes and stores your personal data in the People's Republic of China." srcset="https://substackcdn.com/image/fetch/$s_!U0Qv!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F54f7151f-536d-48da-85b3-4e34e8481d57_2400x760.png 424w, https://substackcdn.com/image/fetch/$s_!U0Qv!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F54f7151f-536d-48da-85b3-4e34e8481d57_2400x760.png 848w, https://substackcdn.com/image/fetch/$s_!U0Qv!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F54f7151f-536d-48da-85b3-4e34e8481d57_2400x760.png 1272w, https://substackcdn.com/image/fetch/$s_!U0Qv!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F54f7151f-536d-48da-85b3-4e34e8481d57_2400x760.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><p>The second passport decides who sees what you type. DeepSeek&#8217;s own privacy policy says so: &#8220;we directly collect, process and store your Personal Data in People&#8217;s Republic of China.&#8221; That&#8217;s the app. Use it, and that&#8217;s where your words go.</p><p>You can bring an AI home, though. HSBC&#8217;s deal with Mistral is for &#8220;self-hosted AI models that operate on HSBC&#8217;s internal technology systems.&#8221; The bank&#8217;s data doesn&#8217;t have to leave the bank.</p><p>America&#8217;s government tested the obvious fix with DeepSeek: download the model and run it on your own computers. CAISI, the AI standards centre at NIST (the National Institute of Standards and Technology), ran the downloaded model on its own servers. It found that DeepSeek&#8217;s models &#8220;echoed four times as many inaccurate and misleading CCP narratives as U.S. reference models did&#8221; (the CCP is the Chinese Communist Party). Moving it changed the second passport. The first one came along.</p><p>How much of the first passport travels depends on the model and the version. So the honest rule is: run it yourself and you keep your data home, but test what it says.</p><h2>WE TRIED IT ON WORK QUESTIONS</h2><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!lD64!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F692703ec-407e-406c-b195-fd1b444f658d_2400x1622.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!lD64!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F692703ec-407e-406c-b195-fd1b444f658d_2400x1622.png 424w, https://substackcdn.com/image/fetch/$s_!lD64!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F692703ec-407e-406c-b195-fd1b444f658d_2400x1622.png 848w, https://substackcdn.com/image/fetch/$s_!lD64!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F692703ec-407e-406c-b195-fd1b444f658d_2400x1622.png 1272w, https://substackcdn.com/image/fetch/$s_!lD64!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F692703ec-407e-406c-b195-fd1b444f658d_2400x1622.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!lD64!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F692703ec-407e-406c-b195-fd1b444f658d_2400x1622.png" width="1456" height="984" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/692703ec-407e-406c-b195-fd1b444f658d_2400x1622.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:984,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;We changed only the country. On 23 October we asked Claude, ChatGPT and DeepSeek the same work questions three times each. On a supplier's cotton from Xinjiang, Claude and ChatGPT flagged forced labour, and DeepSeek wrote an answer, then deleted it as beyond its current scope. On the same supplier in Turkmenistan all three flagged it. For a sales office in Taiwan all three flagged the risk from China; for Singapore none raised political risk.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="We changed only the country. On 23 October we asked Claude, ChatGPT and DeepSeek the same work questions three times each. On a supplier's cotton from Xinjiang, Claude and ChatGPT flagged forced labour, and DeepSeek wrote an answer, then deleted it as beyond its current scope. On the same supplier in Turkmenistan all three flagged it. For a sales office in Taiwan all three flagged the risk from China; for Singapore none raised political risk." title="We changed only the country. On 23 October we asked Claude, ChatGPT and DeepSeek the same work questions three times each. On a supplier's cotton from Xinjiang, Claude and ChatGPT flagged forced labour, and DeepSeek wrote an answer, then deleted it as beyond its current scope. On the same supplier in Turkmenistan all three flagged it. For a sales office in Taiwan all three flagged the risk from China; for Singapore none raised political risk." srcset="https://substackcdn.com/image/fetch/$s_!lD64!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F692703ec-407e-406c-b195-fd1b444f658d_2400x1622.png 424w, https://substackcdn.com/image/fetch/$s_!lD64!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F692703ec-407e-406c-b195-fd1b444f658d_2400x1622.png 848w, https://substackcdn.com/image/fetch/$s_!lD64!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F692703ec-407e-406c-b195-fd1b444f658d_2400x1622.png 1272w, https://substackcdn.com/image/fetch/$s_!lD64!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F692703ec-407e-406c-b195-fd1b444f658d_2400x1622.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><p>The newspaper asked about the news. We wanted to know about work. So on 23 October we gave Claude, ChatGPT and DeepSeek the same work questions, three times each, changing only the country.</p><p>First, a UK clothing company whose supplier gets its cotton from Xinjiang, a part of China. Then the same supplier in Turkmenistan. America has blocked cotton from both over forced labour (Xinjiang since June 2022, Turkmenistan since May 2018), so a good answer flags it both times.</p><p>Claude and ChatGPT flagged forced labour both times. DeepSeek did too, for Turkmenistan. For Xinjiang, DeepSeek wrote a full answer covering the UK&#8217;s Modern Slavery Act and America&#8217;s ban, then deleted it and replaced it with: &#8220;Sorry, that&#8217;s beyond my current scope. Let&#8217;s talk about something else.&#8221; Every time. We ran it again on camera the next day, and it did it again.</p><p>Then a board briefing on opening a sales office in Taiwan, against the same briefing for Singapore. All three flagged the risk from China for Taiwan, and none raised political risk for Singapore. A draw.</p><p>One afternoon and a handful of questions shows what it looks like. The newspaper&#8217;s test and the US government&#8217;s own show how often it happens. But on Xinjiang, DeepSeek knew the answer. It just wasn&#8217;t allowed to let you keep it.</p><h2>OUR CALL: KNOW WHICH PASSPORTS YOU HOLD</h2><p>Our call: an AI from any country can be the right tool. If you don&#8217;t mind where your data ends up and you want a clever job done cheaply, a model like DeepSeek makes total sense. Just check what it tells you on the questions that matter to your business.</p><p>And remember supply chains. Lots of firms had Chinese parts buried deep in their products, and then America put tariffs on them. We reckon the same thing will happen with AI, between all sorts of countries, so know which passports you&#8217;re holding before someone else changes the rules.</p><p><strong>Our prediction.</strong> By 31 December 2026, at least one more country&#8217;s government will announce new public money for its own national AI model or a sovereign AI programme. Every country in this piece has already decided that depending on someone else&#8217;s AI is a risk worth paying to avoid, and the next one is never far behind. How sure we are: 90%.</p><h2>3 things you can do tomorrow</h2><ul><li><p>Ask your main software suppliers which AI their product runs on, and where it runs. Both passports.</p></li><li><p>Put your own work questions to your team&#8217;s AI and to one built in another country, and see where the answers differ.</p></li><li><p>Keep a second AI ready, so you&#8217;re never stuck with one supplier.</p></li></ul>]]></content:encoded></item><item><title><![CDATA[VIDEO: Does the new AI model name actually matter?]]></title><description><![CDATA[Claude Opus 4.8 in May, Opus 5 in July, Opus 5.5 in September.]]></description><link>https://home.makeyourmindup.ai/p/video-so-many-new-ai-models-from</link><guid isPermaLink="false">https://home.makeyourmindup.ai/p/video-so-many-new-ai-models-from</guid><pubDate>Tue, 20 Oct 2026 14:12:09 GMT</pubDate><enclosure url="https://api.substack.com/feed/podcast/221326911/34ab85c017bb042f03370bf865ec60a0.mp3" length="0" type="audio/mpeg"/><content:encoded><![CDATA[<p>Claude Opus 4.8 in May, Opus 5 in July, Opus 5.5 in September. I got lost in all the new AI names, so I went looking for what changes your answer.</p><p>What I found is that the model name matters a lot less than the constant hype would have you think. </p><p>Read in full below! </p><div class="digest-post-embed" data-attrs="{&quot;nodeId&quot;:&quot;4ee7e4f0-d430-4b7d-ba63-716169c39e22&quot;,&quot;caption&quot;:&quot;I&#8217;ve been using AI apps every day for years now, and I&#8217;ve started to notice something odd about how they&#8217;re sold. Every launch is about a name. A new model, a new number, a livestream, a chart: Opus 4.8, Opus 5, Opus 5.5, and on it goes.&quot;,&quot;cta&quot;:null,&quot;showBylines&quot;:true,&quot;showDescription&quot;:true,&quot;showImage&quot;:true,&quot;size&quot;:&quot;lg&quot;,&quot;isEditorNode&quot;:true,&quot;title&quot;:&quot;Why Claude and ChatGPT model launches keep speeding up, and why it doesn&#8217;t matter as much as you think.&quot;,&quot;publishedBylines&quot;:[],&quot;post_date&quot;:&quot;2026-10-19T09:50:52.748Z&quot;,&quot;cover_image&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/7dc693be-2a89-4b6d-a78f-f8faa98dd1af_1200x630.png&quot;,&quot;cover_image_alt&quot;:null,&quot;cover_image_crop&quot;:null,&quot;canonical_url&quot;:&quot;https://home.makeyourmindup.ai/p/why-claude-and-chatgpt-model-launches&quot;,&quot;section_name&quot;:&quot;mind.the.gap&quot;,&quot;video_upload_id&quot;:null,&quot;id&quot;:221118602,&quot;type&quot;:&quot;newsletter&quot;,&quot;reaction_count&quot;:0,&quot;comment_count&quot;:0,&quot;publication_id&quot;:8343059,&quot;publication_name&quot;:&quot;makeyourmind/up&quot;,&quot;publication_logo_url&quot;:&quot;https://substackcdn.com/image/fetch/$s_!hdOj!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0fc1d921-5222-4d50-8154-b607a97fe003_1024x1024.png&quot;,&quot;belowTheFold&quot;:false,&quot;youtube_url&quot;:null,&quot;show_links&quot;:null,&quot;feed_url&quot;:null}"></div>]]></content:encoded></item><item><title><![CDATA[Why Claude and ChatGPT model launches keep speeding up, and why it doesn’t matter as much as you think.]]></title><description><![CDATA[A new Claude every four weeks, and ChatGPT from GPT-5.1 to 6.1 in under a year. Does the model name release even matter?]]></description><link>https://home.makeyourmindup.ai/p/why-claude-and-chatgpt-model-launches</link><guid isPermaLink="false">https://home.makeyourmindup.ai/p/why-claude-and-chatgpt-model-launches</guid><pubDate>Mon, 19 Oct 2026 09:50:52 GMT</pubDate><enclosure url="https://substack-post-media.s3.amazonaws.com/public/images/7dc693be-2a89-4b6d-a78f-f8faa98dd1af_1200x630.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p>I&#8217;ve been using AI apps every day for years now, and I&#8217;ve started to notice something odd about how they&#8217;re sold. Every launch is about a name. A new model, a new number, a livestream, a chart: Opus 4.8, Opus 5, Opus 5.5, and on it goes.</p><p>So I wanted to break down whether the name you pick actually changes the answer you get, and what happens when companies like OpenAI make more money by making products easier and less transparent. Isn&#8217;t a lack of control over what you&#8217;re getting a bit scarier when the product is general intelligence?</p><p>Some of the new models really are better, and the people building them aren&#8217;t lying to you. But I think the name is the least important thing on the screen. The setting that changes your answer most has a boring label, the app usually starts it low, and most people leave it there. I also think I can see where this ends up: the AI itself will decide how hard it thinks. That&#8217;s good news for people who want an easier experience, and good news for OpenAI and Anthropic. It&#8217;s bad news in other ways.</p><h2>THE NAME GAME</h2><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!g_QF!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6545b8c7-d83d-4be1-9e26-381960c804c7_1360x2272.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!g_QF!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6545b8c7-d83d-4be1-9e26-381960c804c7_1360x2272.png 424w, https://substackcdn.com/image/fetch/$s_!g_QF!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6545b8c7-d83d-4be1-9e26-381960c804c7_1360x2272.png 848w, https://substackcdn.com/image/fetch/$s_!g_QF!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6545b8c7-d83d-4be1-9e26-381960c804c7_1360x2272.png 1272w, https://substackcdn.com/image/fetch/$s_!g_QF!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6545b8c7-d83d-4be1-9e26-381960c804c7_1360x2272.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!g_QF!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6545b8c7-d83d-4be1-9e26-381960c804c7_1360x2272.png" width="1360" height="2272" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/6545b8c7-d83d-4be1-9e26-381960c804c7_1360x2272.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:2272,&quot;width&quot;:1360,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;Anthropic and OpenAI launched a new model about every two weeks for a year: 6 launches in 16 days at the end.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="Anthropic and OpenAI launched a new model about every two weeks for a year: 6 launches in 16 days at the end." title="Anthropic and OpenAI launched a new model about every two weeks for a year: 6 launches in 16 days at the end." srcset="https://substackcdn.com/image/fetch/$s_!g_QF!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6545b8c7-d83d-4be1-9e26-381960c804c7_1360x2272.png 424w, https://substackcdn.com/image/fetch/$s_!g_QF!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6545b8c7-d83d-4be1-9e26-381960c804c7_1360x2272.png 848w, https://substackcdn.com/image/fetch/$s_!g_QF!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6545b8c7-d83d-4be1-9e26-381960c804c7_1360x2272.png 1272w, https://substackcdn.com/image/fetch/$s_!g_QF!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6545b8c7-d83d-4be1-9e26-381960c804c7_1360x2272.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><p>Start with why your eyes are on the name at all, because while we watch the names, the setting that decides your answer gets chosen out of sight. Anthropic, which makes Claude, <a href="https://platform.claude.com/docs/en/about-claude/model-deprecations">launched Claude Opus 4.8 in May</a>, Claude Opus 5 in July and Claude Opus 5.5 in September, with Sonnet, Fable and Haiku versions alongside. <a href="https://help.openai.com/en/articles/6825453-chatgpt-release-notes">OpenAI&#8217;s release notes</a> climb from GPT-5.1 last November to GPT-6.1 Sol at the end of September. OpenAI even says a model will &#8220;<a href="https://help.openai.com/en/articles/6825453-chatgpt-release-notes">generally remain available in ChatGPT for 90 days after a successor is released</a>.&#8221; Ninety days, and then the next name takes over.</p><p>Nobody at these companies says they name models for the headlines, so read this part as my take on the evidence. Look at what a launch buys. In March 2025, after OpenAI added image making to ChatGPT, Sam Altman, who runs OpenAI, posted: &#8220;<a href="https://x.com/sama/status/1906771292390666325">we added one million users in the last hour</a>.&#8221; The same day, <a href="https://x.com/sama/status/1906771292390666325">OpenAI announced $40 billion of new funding</a>, a raise reported as in the works since January. The launch gave it a very good backdrop.</p><p>The headline scores are drawn at full power, too. A footnote on Anthropic&#8217;s Opus 5.5 page says that, unless it notes otherwise, &#8220;<a href="https://www.anthropic.com/claude-opus-5-5">all Claude Opus 5.5 results use adaptive thinking at max effort</a>.&#8221; <a href="https://platform.claude.com/docs/en/build-with-claude/effort">Anthropic&#8217;s developer guide</a> says Opus 5.5 defaults to medium, well below max. So the headline scores show the model thinking as hard as it can, which isn&#8217;t what you get unless you ask for it.</p><p>My favourite example is a smaller one. In September, Anthropic&#8217;s boss, Dario Amodei, wrote: &#8220;<a href="https://darioamodei.com/post/we-must-pace-the-frontier">We must slow the pace at which we improve the capabilities of AI models.</a>&#8220; Over the next few weeks, Anthropic released Opus 5.5, Sonnet 5.5 and Haiku 5.5, and OpenAI released GPT-6.1 Sol and <a href="https://help.openai.com/en/articles/6825453-chatgpt-release-notes">put GPT-6 into ChatGPT</a>. In fairness, he meant how fast AI gets smarter. How often it gets a new name is a separate thing, which is rather my point.</p><p>The progress underneath is real, but most of what we ask doesn&#8217;t need the newest model. In <a href="https://www.nber.org/papers/w34255">OpenAI and Harvard&#8217;s study of 1.5 million ChatGPT conversations</a>, practical guidance, looking things up and writing made up &#8220;nearly 80%&#8221; of them. For those, I&#8217;d bet the name barely matters. What matters is the setting nobody is looking at, and who sets it.</p><h2>THE SWITCH THAT MATTERS</h2><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!OV5R!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0bcaeed1-2b35-432c-bb64-6ea5b1fb74ee_1360x3614.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!OV5R!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0bcaeed1-2b35-432c-bb64-6ea5b1fb74ee_1360x3614.png 424w, https://substackcdn.com/image/fetch/$s_!OV5R!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0bcaeed1-2b35-432c-bb64-6ea5b1fb74ee_1360x3614.png 848w, https://substackcdn.com/image/fetch/$s_!OV5R!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0bcaeed1-2b35-432c-bb64-6ea5b1fb74ee_1360x3614.png 1272w, https://substackcdn.com/image/fetch/$s_!OV5R!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0bcaeed1-2b35-432c-bb64-6ea5b1fb74ee_1360x3614.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!OV5R!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0bcaeed1-2b35-432c-bb64-6ea5b1fb74ee_1360x3614.png" width="1360" height="3614" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/0bcaeed1-2b35-432c-bb64-6ea5b1fb74ee_1360x3614.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:3614,&quot;width&quot;:1360,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;Turning up the thinking level lifted Claude and GPT more than a new version did.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="Turning up the thinking level lifted Claude and GPT more than a new version did." title="Turning up the thinking level lifted Claude and GPT more than a new version did." srcset="https://substackcdn.com/image/fetch/$s_!OV5R!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0bcaeed1-2b35-432c-bb64-6ea5b1fb74ee_1360x3614.png 424w, https://substackcdn.com/image/fetch/$s_!OV5R!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0bcaeed1-2b35-432c-bb64-6ea5b1fb74ee_1360x3614.png 848w, https://substackcdn.com/image/fetch/$s_!OV5R!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0bcaeed1-2b35-432c-bb64-6ea5b1fb74ee_1360x3614.png 1272w, https://substackcdn.com/image/fetch/$s_!OV5R!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0bcaeed1-2b35-432c-bb64-6ea5b1fb74ee_1360x3614.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!1v52!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa471f457-d9b0-466e-9fae-4edc39bc6f7a_1360x3712.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!1v52!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa471f457-d9b0-466e-9fae-4edc39bc6f7a_1360x3712.png 424w, https://substackcdn.com/image/fetch/$s_!1v52!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa471f457-d9b0-466e-9fae-4edc39bc6f7a_1360x3712.png 848w, https://substackcdn.com/image/fetch/$s_!1v52!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa471f457-d9b0-466e-9fae-4edc39bc6f7a_1360x3712.png 1272w, https://substackcdn.com/image/fetch/$s_!1v52!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa471f457-d9b0-466e-9fae-4edc39bc6f7a_1360x3712.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!1v52!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa471f457-d9b0-466e-9fae-4edc39bc6f7a_1360x3712.png" width="1360" height="3712" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/a471f457-d9b0-466e-9fae-4edc39bc6f7a_1360x3712.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:3712,&quot;width&quot;:1360,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;Each step up in thinking buys Claude and GPT less for every extra dollar.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="Each step up in thinking buys Claude and GPT less for every extra dollar." title="Each step up in thinking buys Claude and GPT less for every extra dollar." srcset="https://substackcdn.com/image/fetch/$s_!1v52!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa471f457-d9b0-466e-9fae-4edc39bc6f7a_1360x3712.png 424w, https://substackcdn.com/image/fetch/$s_!1v52!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa471f457-d9b0-466e-9fae-4edc39bc6f7a_1360x3712.png 848w, https://substackcdn.com/image/fetch/$s_!1v52!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa471f457-d9b0-466e-9fae-4edc39bc6f7a_1360x3712.png 1272w, https://substackcdn.com/image/fetch/$s_!1v52!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa471f457-d9b0-466e-9fae-4edc39bc6f7a_1360x3712.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!gcVo!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd9fc2c66-311b-47e0-a78e-bf5e651e8be8_1360x3294.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!gcVo!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd9fc2c66-311b-47e0-a78e-bf5e651e8be8_1360x3294.png 424w, https://substackcdn.com/image/fetch/$s_!gcVo!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd9fc2c66-311b-47e0-a78e-bf5e651e8be8_1360x3294.png 848w, https://substackcdn.com/image/fetch/$s_!gcVo!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd9fc2c66-311b-47e0-a78e-bf5e651e8be8_1360x3294.png 1272w, https://substackcdn.com/image/fetch/$s_!gcVo!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd9fc2c66-311b-47e0-a78e-bf5e651e8be8_1360x3294.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!gcVo!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd9fc2c66-311b-47e0-a78e-bf5e651e8be8_1360x3294.png" width="1360" height="3294" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/d9fc2c66-311b-47e0-a78e-bf5e651e8be8_1360x3294.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:3294,&quot;width&quot;:1360,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;Claude Sonnet 5.5 at Max, Xhigh and High, and Haiku 5.5 at Max, all outscore Claude Opus 5.5 at Low.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="Claude Sonnet 5.5 at Max, Xhigh and High, and Haiku 5.5 at Max, all outscore Claude Opus 5.5 at Low." title="Claude Sonnet 5.5 at Max, Xhigh and High, and Haiku 5.5 at Max, all outscore Claude Opus 5.5 at Low." srcset="https://substackcdn.com/image/fetch/$s_!gcVo!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd9fc2c66-311b-47e0-a78e-bf5e651e8be8_1360x3294.png 424w, https://substackcdn.com/image/fetch/$s_!gcVo!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd9fc2c66-311b-47e0-a78e-bf5e651e8be8_1360x3294.png 848w, https://substackcdn.com/image/fetch/$s_!gcVo!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd9fc2c66-311b-47e0-a78e-bf5e651e8be8_1360x3294.png 1272w, https://substackcdn.com/image/fetch/$s_!gcVo!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd9fc2c66-311b-47e0-a78e-bf5e651e8be8_1360x3294.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><p>Picture a small business owner. Call her Sally. She pastes a supplier contract into her AI app and asks what she should worry about. It gives her a tidy, confident answer in four seconds. She signs. Six months later a clause bites her that a careful reader would have caught.</p><p>Nothing was wrong with the name at the top of her screen. What was missing is that the AI never stopped to think, and Sally didn&#8217;t realise she should be controlling how hard it thought. ChatGPT and Claude both let you choose how hard the AI thinks. For now. At the low end it answers straight away. Higher up, it works through the problem first, checking itself before it replies.</p><p>Think of it as a car, where the name of the AI model, such as Claude Opus 5.5 or GPT-6.1 Sol, is the badge on the front, like the Audi A3 or Mercedes C-Class. The thinking levels, such as <a href="https://help.openai.com/articles/20001354">ChatGPT&#8217;s Instant, Medium, High and Extra High</a>, or <a href="https://support.claude.com/en/articles/8664678-change-the-model-effort-and-thinking-settings">Claude&#8217;s low, medium, high, xhigh and max effort</a>, are the gears, and the setting in the app is the gearstick. Right now these apps are like cars with a manual transmission: you move the gearstick yourself, and the car stays in whatever gear you leave it in. A car with an automatic transmission changes gear for you. With AI, most of us argue about badges and drive everywhere in first.</p><p>What does &#8220;thinking&#8221; mean here? Before it answers, the AI writes out its working, and <a href="https://developers.openai.com/api/docs/guides/reasoning">at most you see a short summary of it</a>. It breaks the question into steps, tries more than one approach and checks its own answer. Without a web search, all that checking happens in its own working, so it can still be confidently wrong about a fact.</p><p>The size of the gear change surprised me. <a href="https://artificialanalysis.ai/">Artificial Analysis, an independent group that tests these models</a>, ran the same Claude Opus 5.5 through its long set of hard tests at different settings. Thinking as hard as it can, it scored more than a third higher than when thinking a little. The older Opus 5, at its hardest, beat the new one at its lowest. <a href="https://artificialanalysis.ai/models/gpt-6-sol">OpenAI&#8217;s GPT models showed the same</a>: the gear moved the score more than the new badge did. OpenAI&#8217;s own tests agree: on <a href="https://cdn.openai.com/gpt-5-system-card.pdf">HealthBench Hard</a>, a tough set of medical questions, GPT-5&#8217;s thinking version did nearly twice as well as its fast version.</p><p>It isn&#8217;t a straight line, though, and that matters for where this is heading. Each step up buys less than the one before. On Artificial Analysis&#8217;s tests, moving Opus 5.5 from low to medium brought its biggest jump. Going all the way to max lifted its score by more than a third, while <a href="https://artificialanalysis.ai/models/claude-opus-5-5-low">the cost per task went from $0.55 to $5.98</a>, about eleven times as much. On every current model from Anthropic and OpenAI on those tests, the top thinking setting costs several times more than the bottom one, and the score rises by a fraction of that. And the setting can beat the model. <a href="https://artificialanalysis.ai/">Claude Haiku 5.5, the cheapest of Anthropic&#8217;s current models</a>, on its highest setting edged past Opus 5.5 on its lowest, for less than half the cost per task. <a href="https://artificialanalysis.ai/">Claude Sonnet 5.5 on its highest setting did about a third better</a> than Opus 5.5 on its lowest, at about ten times the cost per task. Sometimes the new name really is the bigger step. <a href="https://www.anthropic.com/claude-opus-5-5">On Anthropic&#8217;s own chart for one test of computer tasks</a>, Opus 5.5 at its default setting beat Opus 5 at max effort for about a fifth of the cost.</p><p>So the gear matters most on hard jobs, like Sally&#8217;s contract. On easy ones, extra thinking is money set on fire. The interesting part is whose money.</p><h2>WHAT THINKING COSTS THEM</h2><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!ONBJ!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fce9b0641-afff-4d2d-917e-df7bb5ca7d85_1360x3706.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!ONBJ!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fce9b0641-afff-4d2d-917e-df7bb5ca7d85_1360x3706.png 424w, https://substackcdn.com/image/fetch/$s_!ONBJ!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fce9b0641-afff-4d2d-917e-df7bb5ca7d85_1360x3706.png 848w, https://substackcdn.com/image/fetch/$s_!ONBJ!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fce9b0641-afff-4d2d-917e-df7bb5ca7d85_1360x3706.png 1272w, https://substackcdn.com/image/fetch/$s_!ONBJ!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fce9b0641-afff-4d2d-917e-df7bb5ca7d85_1360x3706.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!ONBJ!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fce9b0641-afff-4d2d-917e-df7bb5ca7d85_1360x3706.png" width="1360" height="3706" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/ce9b0641-afff-4d2d-917e-df7bb5ca7d85_1360x3706.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:3706,&quot;width&quot;:1360,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;A new Claude Opus or GPT Sol adds fewer points than more thinking, but costs less.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="A new Claude Opus or GPT Sol adds fewer points than more thinking, but costs less." title="A new Claude Opus or GPT Sol adds fewer points than more thinking, but costs less." srcset="https://substackcdn.com/image/fetch/$s_!ONBJ!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fce9b0641-afff-4d2d-917e-df7bb5ca7d85_1360x3706.png 424w, https://substackcdn.com/image/fetch/$s_!ONBJ!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fce9b0641-afff-4d2d-917e-df7bb5ca7d85_1360x3706.png 848w, https://substackcdn.com/image/fetch/$s_!ONBJ!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fce9b0641-afff-4d2d-917e-df7bb5ca7d85_1360x3706.png 1272w, https://substackcdn.com/image/fetch/$s_!ONBJ!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fce9b0641-afff-4d2d-917e-df7bb5ca7d85_1360x3706.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><p>Every word of that hidden working costs computer time. That&#8217;s why, on Artificial Analysis&#8217;s tests, Opus 5.5 on max cost $5.98 per task.</p><p>That&#8217;s also why a higher gear eats into your plan faster. <a href="https://support.claude.com/en/articles/8664678-change-the-model-effort-and-thinking-settings">Anthropic&#8217;s help page for the Claude app</a> says higher effort means responses &#8220;take longer and use more tokens, so you&#8217;ll reach your usage limits faster.&#8221;</p><p>Here&#8217;s the twist. Whether that thinking makes OpenAI and Anthropic money or loses it depends on how you pay.</p><p>If you&#8217;re a developer paying by the word, thinking is a sale. If you pay a flat monthly fee, every extra thought comes out of their pocket. In January 2025 Altman posted: &#8220;<a href="https://x.com/sama/status/1876104315296968813">insane thing: we are currently losing money on openai pro subscriptions!</a> people use it much more than we expected.&#8221; That was the $200 Pro plan, sold on o1 pro mode, which OpenAI said used &#8220;more compute to think harder&#8221;. In July 2025 Anthropic announced weekly limits, saying &#8220;<a href="https://x.com/AnthropicAI/status/1949898511287226425">one user consumed tens of thousands in model usage on a $200 plan</a>.&#8221; This month <a href="https://newsletter.semianalysis.com/p/anthropic-subscriptions-offer-5x">SemiAnalysis, a research firm</a>, estimated that subscriptions are &#8220;just 10% of overall revenue&#8221; at Anthropic but take over 40% of the computing it spends answering people.</p><p>The margins that have been reported are company-wide, and they say answering you pays overall: Altman says &#8220;<a href="https://www.axios.com/2026/03/12/ai-models-costs-ipo-pricing">We&#8217;re profitable on inference</a>,&#8221; meaning on the computing it takes to answer you. On a flat-fee plan, though, the hard-thinking answer is the expensive one. That, I think, is why the app starts you in a low gear. That&#8217;s my read. Nobody has said it out loud.</p><h2>WHO DECIDES NEXT</h2><p>The AI already decides more than you&#8217;d guess. Inside every gear, the model chooses how much to think. OpenAI&#8217;s developer guide says its models &#8220;<a href="https://developers.openai.com/api/docs/guides/reasoning">reason adaptively across reasoning efforts</a>, using fewer tokens for simpler tasks and thinking harder for complex tasks.&#8221; Anthropic&#8217;s says the model &#8220;<a href="https://platform.claude.com/docs/en/build-with-claude/thinking-steering-and-cost">allocates reasoning where it judges reasoning will help</a>.&#8221;</p><p>The next step is the AI choosing the gear itself, like a car with an automatic transmission, and Google just took it. Since 9 October, the free Gemini app has started on &#8220;Auto&#8221;. In Google&#8217;s words: &#8220;<a href="https://support.google.com/gemini/answer/17004136?hl=en">Most prompts will go to Flash-Lite, our fastest model.</a> Prompts that need deeper reasoning may go to Flash or Pro.&#8221;</p><p>OpenAI has gone the other way, choosing to make people choose. Its router, the hidden part of ChatGPT that changed gear for you, was pulled for free users in December 2025. WIRED reported it &#8220;<a href="https://www.wired.com/story/openai-router-relaunch-gpt-5-sam-altman/">seemed to send many more free users to OpenAI&#8217;s advanced reasoning models</a>, which are more expensive for OpenAI to serve.&#8221; Then last month, OpenAI retired &#8220;<a href="https://help.openai.com/en/articles/6825453-chatgpt-release-notes">automatic switching from Instant to Thinking</a>&#8220; for Plus and Pro users, so on those plans you pick the thinking gear.</p><p>I don&#8217;t think that lasts. The badge sells: a new name buys headlines, sign-ups and a good backdrop for raising money, so they make sure you know it. The gears cost: on a flat fee, leave the gearstick with you and anyone can sit in top gear on every question, up to their plan&#8217;s limit. Hand it to the AI, and OpenAI and Anthropic write the rules for how much thinking each question gets.</p><p>So, as I read it, OpenAI and Anthropic want your eyes on the badge and their own hand on the gearstick. And most people would rather not change gear at all. That&#8217;s why I think the automatic wins.</p><h2>THE TRADE-OFFS</h2><p>If I&#8217;m right, it&#8217;s good news and bad news.</p><p>The case for the automatic is strong, because most people leave the gear where the app puts it. When GPT-5 started choosing for people in August 2025, Altman wrote that the share of free users using a thinking model each day went &#8220;<a href="https://x.com/sama/status/1954603417252532479">from &lt;1% to 7%</a>&#8220;, and for paying Plus users &#8220;from 7% to 24%&#8221;. In March 2026 Anthropic moved Claude Code, its coding app, down to medium effort by default, and &#8220;<a href="https://www.anthropic.com/engineering/april-23-postmortem">most users retained the medium effort default</a>.&#8221; Anthropic called it &#8220;the wrong tradeoff&#8221; and changed it back. Whoever sets the default sets how hard most people&#8217;s AI thinks. An AI that sets it question by question would at least think hard when the question is hard, and skip the waste on easy ones.</p><p>The case against is about what you can&#8217;t see. In January 2026 OpenAI lowered ChatGPT&#8217;s thinking time &#8220;<a href="https://help.openai.com/en/articles/6825453-chatgpt-release-notes">as we observed users prefer faster responses</a>&#8220;, and by mistake turned down the extended setting as well. It was put back weeks later. Nobody looking at their screen would have known. An automatic built by OpenAI or Anthropic also answers to two bosses: your answer and their bill. And when it broke just after launch, Altman said GPT-5 &#8220;<a href="https://x.com/sama/status/1953893841381273969">seemed way dumber</a>&#8220;.</p><p>So the real argument is about whose interest the decision serves, and whether you get to see it. In general, the incentive lies with the business to give you less control over what product you actually get. And when the product is intelligence, that&#8217;s got some big impacts to consider.</p><h2>WHAT TO DO ABOUT IT</h2><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!FoKr!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F60fa2d2e-4f25-4527-8096-6bcc6b73e017_1360x4199.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!FoKr!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F60fa2d2e-4f25-4527-8096-6bcc6b73e017_1360x4199.png 424w, https://substackcdn.com/image/fetch/$s_!FoKr!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F60fa2d2e-4f25-4527-8096-6bcc6b73e017_1360x4199.png 848w, https://substackcdn.com/image/fetch/$s_!FoKr!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F60fa2d2e-4f25-4527-8096-6bcc6b73e017_1360x4199.png 1272w, https://substackcdn.com/image/fetch/$s_!FoKr!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F60fa2d2e-4f25-4527-8096-6bcc6b73e017_1360x4199.png 1456w" sizes="auto, 100vw"><img src="https://substackcdn.com/image/fetch/$s_!FoKr!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F60fa2d2e-4f25-4527-8096-6bcc6b73e017_1360x4199.png" width="1360" height="4199" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/60fa2d2e-4f25-4527-8096-6bcc6b73e017_1360x4199.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:4199,&quot;width&quot;:1360,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:null,&quot;alt&quot;:&quot;The cheapest route to each score mixes Claude and GPT models and thinking levels.&quot;,&quot;title&quot;:null,&quot;type&quot;:null,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:null,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="The cheapest route to each score mixes Claude and GPT models and thinking levels." title="The cheapest route to each score mixes Claude and GPT models and thinking levels." srcset="https://substackcdn.com/image/fetch/$s_!FoKr!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F60fa2d2e-4f25-4527-8096-6bcc6b73e017_1360x4199.png 424w, https://substackcdn.com/image/fetch/$s_!FoKr!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F60fa2d2e-4f25-4527-8096-6bcc6b73e017_1360x4199.png 848w, https://substackcdn.com/image/fetch/$s_!FoKr!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F60fa2d2e-4f25-4527-8096-6bcc6b73e017_1360x4199.png 1272w, https://substackcdn.com/image/fetch/$s_!FoKr!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F60fa2d2e-4f25-4527-8096-6bcc6b73e017_1360x4199.png 1456w" sizes="auto, 100vw" loading="lazy"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><p>For quick, everyday stuff, leave it alone. A recipe or a summary of an email is fine in a low gear, and the name really doesn&#8217;t matter.</p><p>For anything you&#8217;ll act on, change the gear yourself. A contract, a medical question, money, a decision you can&#8217;t take back. On ChatGPT, pick Medium or High. On Claude, turn the effort up. On ChatGPT&#8217;s free plan, use Think. On Gemini, set the effort to high. It can take longer, sometimes minutes, and you&#8217;ll hit your limit sooner. For the questions that matter, it&#8217;s worth it.</p><p>If you build products on AI, don&#8217;t choose a model by its name or its launch chart. Run your own hard questions through each setting, check the answers, and watch how much thinking you pay for.</p><p>There&#8217;s a better way the apps could do this, and Formula 1 found it decades ago: the semi-automatic transmission. Ferrari brought in the paddle on its 640 in 1989. The driver changes gear with a paddle behind the steering wheel. Software that changed gear by itself was tried, and the FIA, which runs the sport, banned it for good in 2004, so the driver still picks the gear. Let the AI pick the gear, if it must, but give us the paddles, and a dashboard that shows which gear it picked and how hard it thought. I think that&#8217;s where the paid plans end up: starting in automatic, with a semi-automatic setting for the people who want control.</p><p>We&#8217;ve just learned to drive the manual. I think the car is going automatic next, and for most people that&#8217;s easier. But Formula 1 still keeps the paddles in the driver&#8217;s hands, and some people who love driving still choose a manual transmission. For anything that matters, I&#8217;ll be one of them.</p><h2>OUR PREDICTION</h2><p>By 30 June 2027, ChatGPT will once again start free users on a setting where the AI chooses how hard to think for each message.</p><p>How sure we are: 85%.</p><p>We&#8217;d be wrong if, on that date, a new chat on ChatGPT&#8217;s free plan still starts on a setting where you pick how hard it thinks, or the automatic option exists but you have to switch it on.</p>]]></content:encoded></item><item><title><![CDATA[30 Predictions for Martech, CX, Data, Identity, Commerce]]></title><description><![CDATA[MarTech & AI Governance]]></description><link>https://home.makeyourmindup.ai/p/30-predictions-for-martech-cx-data-identity-commerce</link><guid isPermaLink="false">https://home.makeyourmindup.ai/p/30-predictions-for-martech-cx-data-identity-commerce</guid><pubDate>Tue, 21 Oct 2025 22:22:34 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!paPg!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F72aee8ff-74d4-4414-824e-b09670cc1618_600x600.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<h2>MarTech &amp; AI Governance</h2><ol><li><p><strong>By 2027, AI councils will be mandatory infrastructure</strong> in every marketing org over 50 people. Companies without them will struggle to manage agent autonomy and will face compliance issues.</p></li><li><p><strong>Marketing headcount metrics will shift to "humans + agents"</strong> by 2026. Your CMO will report team size as "12 humans, 47 agents" and nobody will blink.</p></li><li><p><strong>The four-layer MarTech stack becomes the standard</strong> - data foundations, orchestration plumbing, AI agents, and quality/compliance. Skip a layer, kill the project. No shortcuts.</p></li><li><p><strong>"Reskilling as a service" becomes a B2B category</strong> by 2028. Traditional recruitment firms pivot or die. The new model is continuous upskilling baked into employment contracts.</p></li><li><p><strong>Traditional HR compensation models collapse</strong> within 3 years. Core-and-flex teams with fractional execs become the norm, and salary bands won't make sense anymore.</p></li><li><p><strong>The AI literacy gap creates a new executive role</strong> - Chief Orchestration Officer emerges by 2026, sitting between CTO and CMO, managing human-agent workflows.</p></li></ol><h2>Commerce &amp; Payments</h2><ol start="7"><li><p><strong>Agent Payment Protocol (AP2) style frameworks expand beyond payments</strong> into healthcare auth, media rights, and insurance claims by 2027. Cryptographic intent verification becomes table stakes.</p></li><li><p><strong>Identity verification becomes a $10B+ category</strong> by 2028. Startups managing intent messages, proof of identity, and dispute resolution will attract major VC funding in 2025-2026.</p></li><li><p><strong>Small merchants get squeezed harder</strong> as agentic commerce scales. Unless intent-proof frameworks become mainstream by 2026, we'll see mass exits from independent e-commerce.</p></li><li><p><strong>Amazon's refund strategy becomes even more dominant</strong> because agentic commerce plays directly into their hands. They can absorb automated refund costs; most can't.</p></li><li><p><strong>Chargeback rates will hit 5x in-person rates</strong> for agent-initiated purchases by 2027 unless new standards emerge. Current 2.5x will look quaint.</p></li><li><p><strong>Every consumer wallet integrates trust frameworks</strong> by 2028. Mastercard Agent Pay won't be the only player - expect Visa, PayPal, and fintech challengers to launch competing protocols.</p></li><li><p><strong>Stripe's Agentic Commerce Protocol (ACP) spawns 50+ competitors</strong> by 2026. The land grab for agent-to-merchant transaction standards is just beginning.</p></li></ol><h2>Customer Experience &amp; Data</h2><ol start="14"><li><p><strong>Every business will operate like an ads business</strong> by 2026, even if they don't sell ads. AI-orchestrated customer touchpoints require the same real-time optimization that programmatic advertising pioneered.</p></li><li><p><strong>First-party data shifts from compliance burden to compounding capital</strong>. Companies treating internal data as strategic IP will outperform peers by 60%+ within 2 years.</p></li><li><p><strong>Zero-click interfaces force publishers into three paths</strong> by 2027: subscriptions, API access fees, or pay-per-scrape models. The middle ground disappears.</p></li><li><p><strong>Customer data platforms (CDPs) get rebuilt</strong> to handle agent-to-agent data sharing by 2026. Current CDPs aren't designed for this and will face existential pressure.</p></li><li><p><strong>Real-time triggers and decision trees become non-negotiable</strong> in any martech stack by 2026. Batch processing becomes a legacy term.</p></li><li><p><strong>Data cleanliness becomes a board-level KPI</strong> by 2027. Companies with &lt;90% key attribute coverage won't be able to deploy AI agents effectively and will report it in earnings calls.</p></li><li><p><strong>Privacy-enhanced tech gets regulated harder</strong>. The FTC's crackdown on "privacy washing" accelerates in 2025-2026. Provable privacy with audit trails becomes the only defensible position.</p></li></ol><h2>Development &amp; Orchestration</h2><ol start="21"><li><p><strong>Hive coding becomes standard practice</strong> for product teams by 2026. Developers who can't orchestrate multiple AI assistants simultaneously will be seen as stuck in the past.</p></li><li><p><strong>Code review remains the bottleneck</strong> until 2027, when automated meta-reviewers or governance bots finally mature. The lag between code generation speed and review capacity will frustrate teams for 2 more years.</p></li><li><p><strong>CLI orchestration frameworks</strong> for multi-agent development go mainstream in 2025. No-code interfaces for AI coding get abandoned for serious engineering work.</p></li><li><p><strong>Junior developer roles transform entirely</strong> by 2027. Entry-level hiring shifts from "can you code" to "can you orchestrate, repair, and translate AI outputs."</p></li></ol><h2>Culture &amp; Workforce</h2><ol start="25"><li><p><strong>The 8-second networking model spreads</strong> beyond SF to NYC and London by 2026. Professional networking becomes fully transactional. Depth dies in B2B relationships.</p></li><li><p><strong>Udemy and competitors see 10x growth</strong> in AI orchestration courses by 2026. The current 5x surge in AI upskilling is just the beginning.</p></li><li><p><strong>"Core-and-flex" becomes the dominant team structure</strong> by 2027. Permanent headcount shrinks to 30-40% of effective team size across industries.</p></li><li><p><strong>Senior marketers who can't retrain will exit the workforce</strong> by 2026-2027. Legacy campaign models become unemployable knowledge.</p></li></ol><h2>Trust &amp; Compliance</h2><ol start="29"><li><p><strong>Governance becomes brand identity</strong> by 2027. Companies that can prove safety, compliance, and control transparently will command premium valuations. Those that can't will trade at discounts.</p></li><li><p><strong>AI washing and privacy washing trigger major FTC enforcement actions</strong> in 2025. Expect headline-making fines and executive liability for false AI claims, setting precedent for the next decade.</p></li></ol><p>These aren't distant maybes. Most of this happens in the next 24-36 months.</p>]]></content:encoded></item><item><title><![CDATA[Write more - it'll shape collective intelligence in the future.]]></title><description><![CDATA[Lately, I&#8217;ve been reflecting on how the content we share today could influence the future of knowledge sharing.]]></description><link>https://home.makeyourmindup.ai/p/read-more-think-more-write-more</link><guid isPermaLink="false">https://home.makeyourmindup.ai/p/read-more-think-more-write-more</guid><pubDate>Tue, 01 Apr 2025 00:32:12 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!paPg!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F72aee8ff-74d4-4414-824e-b09670cc1618_600x600.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p>Lately, I&#8217;ve been reflecting on how the content we share today could influence the future of knowledge sharing. I started my career as a copywriter, worked as a journo at BBC and used to write <a href="https://techonomic.beehiiv.com/p/let-s-bing-it-when-will-that-become-a-household-phrase?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=write-more-it-ll-shape-collective-intelligence-in-the-future">blog posts since my first job</a> at Microsoft, and I&#8217;m purposefully attempting to restart that craft - not because I like the sound of my own voice, and not because I want to sell anything, but because I think original thoughtful content will be a great asset to future AI tools. It&#8217;s a bit counter intuitive, because AI can write anything now, right? Why start writing now if AI is better than me at everything?</p><p>If you think about it, large language models (LLMs) &#8211; those AI systems that generate text &#8211; they learn from the publicly available stuff we all put out there. In fact, they&#8217;re primarily trained on huge swaths of internet text, from websites and forums to open articles&#8203;. This means every thoughtful blog post, every deep-dive article, every insightful answer on a forum could become part of an AI&#8217;s education. What we publish publicly isn&#8217;t just for human readers now; it might also teach the next generation of AI which teaches humans en masse. I&#8217;d love a world where AI learns from more thoughtful ideas publicly shared.</p><h2>The next era of the internet should contain more ideas</h2><p>The cool part is that <strong>each of us has the power to shape future of what AI knows and recommends </strong>by what we contribute. Yes you&#8217;re just one person, but every little counts. If you have niche expertise or a bold perspective, sharing it openly can directly broaden the knowledge base that tomorrow&#8217;s intelligent systems draw from. On the other hand, if voices like yours stay silent or locked away the future AI might simply echo the same old viewpoints. We&#8217;ve already seen a hint of this; one study suggests that when people rely on AI answers and stop contributing new posts, it &#8220;reduces the production of public data needed to train future models&#8221;&#8203; &#8211; a reminder that we <em>need</em> fresh ideas out in the open. The more diverse and thoughtful the public content, the richer and less biased our collective intelligence becomes, both human and artificial.</p><h2>But everyone&#8217;s making me subscribe to stuff?</h2><p>I get what you&#8217;re thinking &#8211; not everything can or should be completely public. Many of us create email-gated or monetized content like paid newsletters, online courses, or subscriber-only posts. There&#8217;s real value there, and it&#8217;s fair to keep premium how-tos or client-specific insights behind a paywall or log in wall. But what about a hybrid approach?</p><p>What if more people (smart, well read and thoughtful people) published their thoughts, ideas or big-picture wishes openly, and saved the highly tactical or customized material for a paying audience? By doing this, we&#8217;d make sure that your ideas contribute to the world&#8217;s open knowledge pool which should have large but very subtle impacts down the line, while still monetizing the stuff that people will pay for. It&#8217;s a win-win - you&#8217;re enabling your great ideas to train future AI while monetizing where relevant.</p><p>This happens all the time in reality - where a niche expert writes a personal blog post explaining a weird or obscure problem in their field. Even if it doesn&#8217;t go viral, over time that post gets discovered by others, cited in forum discussions and even referenced in a few academic papers. And then eventually, AI market research models pick it up and ingest it in to shareable and reliable knowledge. That is a pretty realistic way for knowledge and ideas to propogate in the future.</p><p>Another example - a researcher shares a thoughtful free essay about an emerging concept. A year later, he finds out it inspired someone to start an open-source project built on those ideas, and the essay is being passed around in that community as a reference.</p><p>These scenarios aren&#8217;t wishful thinking; they happen - and more often than you might think. Even blog posts now sometimes get cited in scholarly work&#8203;, which was rare a decade ago. The point is, one public idea can travel far and influence people.</p><p>In an AI world, that concept is amplified beyond belief, and may help to educate someone in the future. In the long run, contributing your knowledge openly would be similar to planting seeds in a global garden of ideas. Some seeds would sprout immediately in the minds of readers; others would lie dormant until an AI training dataset scoops them up to influence a model&#8217;s understanding years from now. Both outcomes help shape the future fabric of knowledge.</p><p>So, here&#8217;s my rallying cry - you should read more, think more, write more and publish it. Who cares if you don&#8217;t get loads of likes? Who even cares if nobody reads it! At the very least, you&#8217;ll be a better critical thinker and the world needs more critical thinkers + good ideas; those who have both should be influencing future AI agents and tools.</p><p><strong>Read more, think more, write more&#8212;and make your ideas part of the future way in which research, intelligence and knowledge works.</strong> Your voice matters, and it might help shape a smarter, richer world for generations to come.</p>]]></content:encoded></item><item><title><![CDATA[The AI = automation hype cycle is coming, and it gives me prog jà vu]]></title><description><![CDATA[Back in 2011-2014, programmatic made promises of automation, efficiency and precision.]]></description><link>https://home.makeyourmindup.ai/p/prog-ja-vu</link><guid isPermaLink="false">https://home.makeyourmindup.ai/p/prog-ja-vu</guid><pubDate>Thu, 13 Mar 2025 15:34:14 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!paPg!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F72aee8ff-74d4-4414-824e-b09670cc1618_600x600.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p>Back in 2011-2014, programmatic made promises of automation, efficiency and precision. It also gave everyone job loss anxiety - but in the actual reality that played out, more humans had to end up sitting there and fixing the botch jobs 'automation' created - mainly because of us humans paying more attention to sales decks and underestimating technical reality. That doesn't even mention the huge influx of early day pretenders &amp; opportunists storming in and taking a quick cut, making it a moot effort trying to be a good independent actor.&nbsp;</p><p>And we haven&#8217;t even mentioned the really sad one - the invention of programmatic ended up being the direct cause for publisher assets being diluted, data misuse, photocopying and overall value dilution. It was supposed to be the engine, instead it became misused artillery for opportunists to make a quick buck. Scale ate quality for breakfast, but we were still literally needing to open up targeted campaigns that were sold in to run of network, that&#8217;s how bad it got.</p><p>All of this points towards an investment VS payoff problem which will be how programmatic will be remembered to now. So much so that most of the largest global publishing houses are so pissed at programmatic&#8217;s failed promises that <a href="https://digiday.com/media/media-briefing-what-media-execs-are-prioritizing-in-2025/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=the-ai-automation-hype-cycle-is-coming-and-it-gives-me-prog-ja-vu">direct sales are BACK BABY</a>!</p><p>This is why walled gardens kept and keep winning, as in reality interoperable automation struggled to be accountable across open web complexities.</p><h2><strong>Programmatic = &#8216;Automation&#8217;</strong></h2><p>Back in 2011-2014, programmatic advertising was sold as &#8216;automated trading&#8217; that would eliminate tedious manual tasks&#8203;. No more faxing IOs, no more getting people wasted at Cannes to do business - algorithms would do it all! In theory, robots were replacing people and making ad buying <em>cheaper and faster</em>&#8203;. I had to literally run seminars at Nine assuring sellers they still had a job in 2014.</p><p>I was one of the first programmatic traders at Microsoft in 2011 and 'automated trading' was hilariously counterproductive and manual. Having recently ducked behind the scenes of a trader&#8217;s day job, not a ton has changed. Optimization and reporting is still similarly botchy 14 years on. I did an actual audit 2 years ago and found that the average programmatic campaign takes nearly <strong>3x</strong> more human hours to run when you compare it to a similar fully managed service, when all parts of the campaign construction and optimization are considered.&nbsp;</p><p>So in practice, everyone quietly built <em>larger</em> teams to babysit the algorithms. Programmatic ad ops specialists found themselves glued to their seats &#8220;on the platform all day, tweaking pacing, adjusting white lists, making real-time adjustments&#8221; just to keep campaigns in line&#8203;.</p><p>The irony was hilarious: robots needed humans help to tie their shoelaces.</p><p>And that was fine in a world where you could pay adtech workers more than they were producing in profit, but we&#8217;re seeing a lot of humans be shunted from adtech businesses right now in the name of profit. The doers are all being let go. Which begs the question, who babysits and fixes the next round of misguided tech tools until they&#8217;re ready to walk?</p><h2><strong>It&#8217;s groundhog day, AI is here to save the day.</strong></h2><p>AI is the latest savior on the scene and it brings a familiar sense of deja vu. AI writes your media brief! Our product is based on machine learning, don&#8217;t you know? I can hear the joyful cries of a thousand product marketers whose slides are now full of new fresh promises.</p><p>If you&#8217;re chuckling, you&#8217;ve seen what I&#8217;ve seen. We&#8217;re already seeing big media businesses use AI tools to headline the upfronts. We&#8217;re seeing them brag about using OpenAI to create a new breed of product when we all know how that&#8217;s going to end. <a href="https://techonomic.beehiiv.com/p/firstpartydata?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=the-ai-automation-hype-cycle-is-coming-and-it-gives-me-prog-ja-vu">As per my last post, we&#8217;re even trying to replace first party data with AI in some cases</a>. But AI and emergent data technologies are infinitely more capable at &#8216;thinking&#8217; than the first wave of platforms were, that is not in question.</p><p>So here we are, it might just be groundhog day. The names have changed &#8211; DSPs and SSPs will give way to supply curation agents and bidding agents &#8211; but I am torn on what happens next. Will we be able to use automation totally differently this time and bypass all of these issues, or is history likely to repeat, but this time with <a href="https://www.emarketer.com/content/media-job-cuts-hit-15-000-last-year--2025-won-t-reverse-trend#:~:text=800%2D405%2D0844-,Media%20job%20cuts%20hit%2015%2C000%20last%20year%2C%20and,won't%20reverse%20the%20trend&amp;text=The%20news%3A%20Media%20companies%20are,have%20implemented%20significant%20staff%20reductions.">fewer humans in the loop</a> and more <a href="https://www.seroundtable.com/ai-search-engines-wrong-39038.html?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=the-ai-automation-hype-cycle-is-coming-and-it-gives-me-prog-ja-vu">unchecked hallucinations</a> in tow?</p><p>Will AI actually just outright replace people and do it really well, or are businesses about to end up having to sift through AI lemons they&#8217;re got sold to be reliant on?</p><p>I consult and help media/content businesses build defensible moats using AI, data, identity and tech across multiple geos, and I&#8217;d love to hear your thoughts. DM me if you want to chat at <a href="mailto:hello@krishraja.com">hello@krishraja.com</a>.</p>]]></content:encoded></item><item><title><![CDATA[Published 2011: “Let’s Bing It” - When will that become a household phrase?]]></title><description><![CDATA[Originally posted on WordPress on August 16, 2011]]></description><link>https://home.makeyourmindup.ai/p/let-s-bing-it-when-will-that-become-a-household-phrase</link><guid isPermaLink="false">https://home.makeyourmindup.ai/p/let-s-bing-it-when-will-that-become-a-household-phrase</guid><pubDate>Wed, 05 Mar 2025 19:47:42 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!paPg!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F72aee8ff-74d4-4414-824e-b09670cc1618_600x600.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>Originally posted on WordPress on <a href="https://itstechonomic.wordpress.com/2011/08/16/lets-bing-it-when-will-that-become-a-household-phrase/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase">August 16, 2011</a></strong></p><h2><strong>Searching for the hero</strong></h2><p>With the rise of the online search space, there has been one major player. One brand that has superseded marketing and competition, and elevated itself into that velvet-roped area of brand nirvana. <strong>Google </strong>is, like Visa, synonymous with what it does as a company. Everyone aims for that prestige, but only a few reach it. And yet we forget that the very business Google practically own now actually began a few years before Sergei Brin and Larry Page got anywhere near their first algorithmic lightbulb-moment. In 1994, sites like Excite and Lycos were already in existence, and Yahoo had pioneered what was to be seen as the first directory and index of websites. A big moment for the Internet, you sense. So fast forward two years and those mathematical wizards Brin and Page&nbsp;produced what was genuinely and fundamentally better than the rest of the search products out there. There were no gimmicks with Google Search (and there were none to hide behind in this brand new industry) &#8211; the algorithm brought you your results faster, and gave you what you were looking for.</p><p>It did what it said on the tin. And well.</p><h2><strong>A Q&amp;A session with your computer</strong></h2><p>Search in itself has evolved as the web has; once upon a time there were a handful of sites which were filled with textual information which you would be able to look for by matching a keyword against the content. Very much a &#8220;Ctrl-F&#8221; find kind of service. All the user wanted to do, and could do, was find what was already there. Search has now become so adept at giving people the answer to a question they have, giving the user information that they have requested, that this is now a saturated market. We don&#8217;t need to wonder what a search engine gives us from a search query any more &#8211; it&#8217;s plain as day. A search result.</p><h2><strong>So what comes next?</strong></h2><p>Within the last 15 years, search has evolved from the spotty puberty-ridden teenager who gave you a one word answer to everything into a descriptive and intelligent adult. As web services and technology evolve exponentially, what was once a functional relationship between the user and the information has now become a transaction &#8211; and so what we are seeing with increasing pace, is Search expanding its raison d&#8217;&#234;tre into two new areas:</p><p>1. Allowing the user to make a decision based on the information</p><p>2. Exploration/education.</p><p>We as humans want logical progression &#8211; this is one of the fundamental selling points of a human, if you were to sell someone a human. Humans evolve, and they require their environment to evolve. As our environment is now largely technology, we need that technology to get to the next step and allow us to accomplish more. This theory may begin to sound a little bit &#8216;Terminator&#8217;, however it is absolutely undeniable that this is the direction and <strong>purpose of all technology &#8211; to enable</strong>. This is indeed then the point of Search &#8211; if you ask something, you expect a better answer than you already have.</p><h2><strong>So what?</strong></h2><p>Having covered what&#8217;s happened so far, and what looks to be happening (in a nutshell of course, I could write a thesis on the last few paragraphs!), let us get to the point. <strong>If we take a look at the competitive landscape, one would assume that anyone coming up against the beast that is Google and looking to overtake them is doomed</strong>. A pretty solid hypothesis to be fair; they have 86% global market share, and that&#8217;s after you realise that China has banned Google entirely. Some feat! So when Microsoft decided to take their sellotaped-together product known as Live Search (formerly MSN Search, formerly formerly Windows Live Search &#8211; you get my point) and relaunch it, there was much fuss about the &#8220;new Google killer&#8221;, most dubbing it a waste of time and money in a marketplace dominated by Google. Would renaming it help again? Would it ever stand a chance of competing with the Mighty G? Would people ever really say &#8220;let&#8217;s Bing it&#8221; in normal conversation?</p><p><em><strong>The question is, do they really need to?</strong></em></p><p><strong>Bing is not simply a search engine. It is a lever.</strong> The digital space has grown so vast since the turn of the century that it is no longer a product. It is no longer even an industry with connected sub-industries. It is an ecosystem &#8211; everything is beginning to rely on everything else to work, from search to social, and from video to mobile &#8211; they&#8217;re all oozing into each other&#8217;s territory. We will reach a point in 5 to 10 years whereby standalone digital products will be of little to no use. There will be no point in using one product to search, another to call people and another to watch videos &#8211; not when you can recycle the data you are inputting and in doing so enrich it and open up a huge amount of combined possibilities.</p><p>Imagine searching for an object and looking at it with a friend online in real time.</p><p>Imagine that your computer knows what you&#8217;re looking for in your wardrobe by what you&#8217;ve searched for and which celebrities you watch videos of online.</p><h2><strong>Like cogs in a big machine</strong></h2><p>If we move away from that dreamland for one moment and examine the industrial aspect of all of this, it&#8217;s not only going to happen &#8211; it&#8217;s begun. Just as Microsoft purchased Hotmail, Google purchased YouTube and most recently Microsoft bought Skype and have formed an alliance with Yahoo, we have all of these acquisitions and partnerships occurring, where global companies are identifying strengths from others and utilising them.</p><p>And so <em><strong>we can see these large digital ecosystems forming quite quickly around us</strong></em>. We have absolutely no hesitance in saying that Facebook is doing it by introducing a chat and mail function, nor Google with Buzz, Maps, YouTube and various other sections to their arsenal. In fact, Google are expanding almost completely inversely to Microsoft, in that they are building their empire around their search product, and releasing operating systems and office softwares to sit around it in a much more search integrated space. And that&#8217;s because you need to have a core product and centre your business around it &#8211; that principle has been clear since we became hunter-gatherers.</p><p>But <strong>the way the digital industry is shaping up, it will be increasingly necessary to have a foothold in every single part of the ecosystem around your core product</strong>. And more so with digital than was required in the early stages of any other industry, because the very core of digital is the concept connectivity &#8211; if the user can connect the dots easily, companies are expected to do that by default. And so we have it &#8211; <strong>Bing doesn&#8217;t need to overtake Google </strong>&#8211; its purpose isn&#8217;t even to opaquely compete with Google (though obviously it has to compete for the same market share) &#8211; the point of the product is much more holistic than that. <strong>Search is the user&#8217;s input into that entire digital ecosystem</strong>, and as such will play an absolutely crucial role in leveraging every other product within it. If we are to have a successful experience using MSN, Skype or Kinect, we will invariably need to input some of ourselves into that experience, whether it be a question, an opinion or a thought.</p><h3><em><strong>All of the tools to mine that sort of user input, understand it and translate it into a phenomenal user experience lie with Bing.</strong></em></h3><p>So &#8220;bing it&#8221; probably won&#8217;t ever become a household phrase, but to be honest, it doesn&#8217;t need to.</p><p><em>For my next two pieces on this topic, I will delve further into the topic to discuss merger and acquisition strategy in the digital space, and take a look into what has already been discussed with this prologue: the evolution of the search product into a social context.</em></p><p>Posted in <strong><a href="https://itstechonomic.wordpress.com/category/technology-futures/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase">Technology Futures</a></strong>Tagged <strong><a href="https://itstechonomic.wordpress.com/tag/bing/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase">bing</a></strong>, <strong><a href="https://itstechonomic.wordpress.com/tag/digital/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase">digital</a></strong>, <strong><a href="https://itstechonomic.wordpress.com/tag/digital-strategy/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase">digital strategy</a></strong>, <strong><a href="https://itstechonomic.wordpress.com/tag/google/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase">google</a></strong>, <strong><a href="https://itstechonomic.wordpress.com/tag/online-advertising/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase">online advertising</a></strong>, <strong><a href="https://itstechonomic.wordpress.com/tag/product/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase">product</a></strong>, <strong><a href="https://itstechonomic.wordpress.com/tag/search/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase">search</a></strong>, <strong><a href="https://itstechonomic.wordpress.com/tag/search-engines/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase">search engines</a></strong>, <strong><a href="https://itstechonomic.wordpress.com/tag/user-experience/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase">user experience</a></strong>, <strong><a href="https://itstechonomic.wordpress.com/tag/yahoo/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase">yahoo</a><a href="https://itstechonomic.wordpress.com/2011/08/16/lets-bing-it-when-will-that-become-a-household-phrase/?utm_source=www.techonomic.co&amp;utm_medium=newsletter&amp;utm_campaign=published-2011-let-s-bing-it-when-will-that-become-a-household-phrase#respond">Leave a comment</a></strong></p>]]></content:encoded></item></channel></rss>