{"id":47610,"date":"2026-04-02T13:16:29","date_gmt":"2026-04-02T11:16:29","guid":{"rendered":"https:\/\/creaskill.ch\/?p=47610"},"modified":"2026-07-27T14:07:38","modified_gmt":"2026-07-27T12:07:38","slug":"kiks01","status":"publish","type":"post","link":"https:\/\/creaskill.ch\/en\/kiks01\/","title":{"rendered":"The cognitive chasm: When AI overtakes our thinking"},"content":{"rendered":"<h4 class=\"gb-text\">2 April 2026 Heinz S\u00fcess<\/h4>\n\n\n\n<!-- creaSKILL Blog-Snippet \u00b7 Breite 1140px \u00b7 ohne Header\/Footer\/Titel -->\n<div class=\"cs-article\">\n<style>\n  @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Serif+Display:ital@0;1&family=DM+Sans:wght@300;400;500;600&display=swap');\n\n  .cs-article {\n    --dunkelblau: #2D2A66;\n    --mittelblau: #167EC1;\n    --hellblau: #78C2EB;\n    --akzent: #C1671A;\n    --ink: #2D2A66;\n    --text: #2e2c3f;\n    --ink-muted: #6d6a82;\n    --paper: #ffffff;\n    --accent-light: #f6ece2;\n    --blue-light-bg: #eaf4fb;\n    --rule: #d8e1ea;\n    --serif: 'DM Serif Display', Georgia, serif;\n    --sans: 'DM Sans', system-ui, sans-serif;\n\n    max-width: 1140px;\n    margin: 0 auto;\n    font-family: var(--sans);\n    background: var(--paper);\n    color: var(--text);\n    line-height: 1.7;\n    font-size: 17px;\n    font-weight: 300;\n    box-sizing: border-box;\n  }\n  .cs-article *, .cs-article *::before, .cs-article *::after { box-sizing: border-box; }\n\n  .cs-article .cs-lead {\n    max-width: 100%;\n    margin: 0 auto;\n    padding: 2.5rem 2rem 1rem;\n  }\n  .cs-article .cs-kicker {\n    font-size: 11px;\n    font-weight: 600;\n    letter-spacing: 0.18em;\n    text-transform: uppercase;\n    color: var(--akzent);\n    margin-bottom: 1.2rem;\n  }\n  .cs-article .cs-deck {\n    font-size: 1.25rem;\n    line-height: 1.6;\n    color: var(--ink);\n    font-weight: 400;\n    max-width: 100%;\n    margin-bottom: 1.8rem;\n    border-left: 3px solid var(--akzent);\n    padding-left: 1.2rem;\n  }\n  .cs-article .cs-byline {\n    font-size: 12px;\n    letter-spacing: 0.08em;\n    text-transform: uppercase;\n    color: var(--ink-muted);\n    display: flex;\n    align-items: center;\n    gap: 1rem;\n    font-weight: 500;\n  }\n  .cs-article .cs-byline::before {\n    content: '';\n    display: block;\n    width: 32px;\n    height: 1px;\n    background: var(--ink-muted);\n  }\n\n  .cs-article .cs-stat-banner {\n    background: var(--dunkelblau);\n    color: var(--paper);\n    padding: 3rem 2rem;\n    margin: 2.5rem 0 0;\n    border-radius: 6px;\n  }\n  .cs-article .cs-stat-banner-inner {\n    max-width: 100%;\n    margin: 0 auto;\n    display: grid;\n    grid-template-columns: repeat(3, 1fr);\n    gap: 2rem;\n    text-align: center;\n  }\n  .cs-article .cs-stat-number {\n    font-family: var(--serif);\n    font-size: 3.4rem;\n    color: var(--hellblau);\n    line-height: 1;\n    margin-bottom: 0.5rem;\n  }\n  .cs-article .cs-stat-label {\n    font-size: 12px;\n    letter-spacing: 0.1em;\n    text-transform: uppercase;\n    color: rgba(255,255,255,0.65);\n    line-height: 1.4;\n  }\n\n  .cs-article .cs-body {\n    max-width: 100%;\n    margin: 0 auto;\n    padding: 0 2rem;\n  }\n  .cs-article h2 {\n    font-family: var(--serif);\n    font-size: 1.9rem;\n    font-weight: 400;\n    color: var(--dunkelblau);\n    margin: 3.5rem 0 1rem;\n    line-height: 1.25;\n  }\n  .cs-article p { margin-bottom: 1.4rem; color: var(--text); }\n  .cs-article em { font-style: italic; }\n  .cs-article strong { font-weight: 600; color: var(--dunkelblau); }\n\n  .cs-article .cs-pullquote {\n    font-family: var(--serif);\n    font-size: 1.6rem;\n    font-style: italic;\n    line-height: 1.35;\n    color: var(--dunkelblau);\n    border-top: 2px solid var(--mittelblau);\n    border-bottom: 2px solid var(--mittelblau);\n    padding: 1.8rem 0;\n    margin: 3rem -1rem;\n    text-align: center;\n  }\n  .cs-article .cs-pullquote cite {\n    display: block;\n    font-family: var(--sans);\n    font-size: 11px;\n    font-style: normal;\n    letter-spacing: 0.12em;\n    text-transform: uppercase;\n    color: var(--mittelblau);\n    margin-top: 1rem;\n    font-weight: 600;\n  }\n\n  .cs-article .cs-loop-box {\n    background: var(--blue-light-bg);\n    border-left: 4px solid var(--mittelblau);\n    padding: 2rem 2rem 2rem 1.8rem;\n    margin: 2.5rem 0;\n    border-radius: 0 6px 6px 0;\n  }\n  .cs-article .cs-loop-box h3 {\n    font-family: var(--serif);\n    font-size: 1.2rem;\n    font-weight: 400;\n    margin-bottom: 1rem;\n    color: var(--mittelblau);\n  }\n  .cs-article .cs-loop-steps { list-style: none; counter-reset: steps; padding: 0; margin: 0; }\n  .cs-article .cs-loop-steps li {\n    counter-increment: steps;\n    display: grid;\n    grid-template-columns: 2rem 1fr;\n    gap: 0.6rem;\n    margin-bottom: 0.9rem;\n    font-size: 15px;\n    line-height: 1.5;\n    color: var(--text);\n  }\n  .cs-article .cs-loop-steps li::before {\n    content: counter(steps);\n    font-family: var(--serif);\n    font-size: 1.2rem;\n    color: var(--akzent);\n    font-weight: 400;\n    line-height: 1.3;\n  }\n\n  .cs-article .cs-timeline {\n    margin: 2rem 0;\n    border-left: 2px solid var(--rule);\n    padding-left: 1.8rem;\n    list-style: none;\n  }\n  .cs-article .cs-timeline-item { position: relative; margin-bottom: 1.6rem; }\n  .cs-article .cs-timeline-item::before {\n    content: '';\n    position: absolute;\n    left: -2.3rem;\n    top: 6px;\n    width: 10px;\n    height: 10px;\n    border-radius: 50%;\n    background: var(--paper);\n    border: 2px solid var(--mittelblau);\n  }\n  .cs-article .cs-timeline-item.cs-highlight::before {\n    background: var(--akzent);\n    border-color: var(--akzent);\n  }\n  .cs-article .cs-timeline-year {\n    font-size: 11px;\n    font-weight: 600;\n    letter-spacing: 0.1em;\n    text-transform: uppercase;\n    color: var(--mittelblau);\n    margin-bottom: 0.3rem;\n  }\n  .cs-article .cs-timeline-item.cs-highlight .cs-timeline-year { color: var(--akzent); }\n  .cs-article .cs-timeline-content { font-size: 15px; color: var(--text); line-height: 1.5; }\n\n  .cs-article .cs-implications {\n    display: grid;\n    grid-template-columns: repeat(2, 1fr);\n    gap: 1.2rem;\n    margin: 2rem 0;\n  }\n  .cs-article .cs-impl-card {\n    border: 1px solid var(--rule);\n    border-top: 3px solid var(--hellblau);\n    padding: 1.4rem;\n    border-radius: 4px;\n  }\n  .cs-article .cs-impl-icon { font-size: 1.4rem; margin-bottom: 0.6rem; display: block; }\n  .cs-article .cs-impl-title {\n    font-family: var(--serif);\n    font-size: 1.05rem;\n    margin-bottom: 0.5rem;\n    font-weight: 400;\n    color: var(--dunkelblau);\n  }\n  .cs-article .cs-impl-text { font-size: 14px; color: var(--ink-muted); line-height: 1.5; margin: 0; }\n\n  .cs-article .cs-closing-box {\n    background: var(--dunkelblau);\n    color: var(--paper);\n    padding: 3rem;\n    margin: 4rem 0 1rem;\n    border-radius: 6px;\n  }\n  .cs-article .cs-closing-box p {\n    color: rgba(255,255,255,0.88);\n    font-size: 1.05rem;\n    line-height: 1.75;\n    margin-bottom: 1rem;\n  }\n  .cs-article .cs-closing-box p:last-child { margin-bottom: 0; }\n  .cs-article .cs-closing-box strong { color: var(--hellblau); }\n\n  .cs-article .cs-source {\n    max-width: 100%;\n    margin: 1.5rem auto 0;\n    padding: 1rem 2rem 0;\n    font-size: 12px;\n    color: var(--ink-muted);\n    font-style: italic;\n  }\n\n  @media (max-width: 600px) {\n    .cs-article .cs-stat-banner-inner { grid-template-columns: 1fr; gap: 1.5rem; }\n    .cs-article .cs-implications { grid-template-columns: 1fr; }\n    .cs-article .cs-pullquote { margin: 2rem 0; }\n    .cs-article .cs-closing-box { padding: 2rem; }\n  }\n<\/style>\n\n  <div class=\"cs-lead\">\n    <p class=\"cs-kicker\">New research \u00b7 March 2026<\/p>\n    <p class=\"cs-deck\">One researcher has measured what many have suspected: AI systems now process information more than a thousand times more efficiently than humans \u2013 and the gap is growing rapidly. Worse still: We are helping to contribute to this.<\/p>\n    <p class=\"cs-byline\">Based on: Eliav, N. \u00b7 MHIL Preprint \u00b7 arXiv:2603.26707<\/p>\n  <\/div>\n\n  <div class=\"cs-stat-banner\">\n    <div class=\"cs-stat-banner-inner\">\n      <div>\n        <div class=\"cs-stat-number\">3.906\u00d7<\/div>\n        <div class=\"cs-stat-label\">Growth of AI context windows since 2017<\/div>\n      <\/div>\n      <div>\n        <div class=\"cs-stat-number\">\u221287%<\/div>\n        <div class=\"cs-stat-label\">Decrease in human capacity to absorb information since 2004<\/div>\n      <\/div>\n      <div>\n        <div class=\"cs-stat-number\">1.111\u00d7<\/div>\n        <div class=\"cs-stat-label\">Raw ratio of AI to humans in 2026<\/div>\n      <\/div>\n    <\/div>\n  <\/div>\n\n  <div class=\"cs-body\">\n    <h2>The problem with the short sentence<\/h2>\n    <p>Imagine it's 2026, you open your AI assistant and type: <em>\u201eWrite me a cancellation for this meeting, two sentences.\u00bb<\/em> Thirty seconds of uninterrupted thought \u2013 that would have cost this task earlier. Today you delegate it without thinking. It has become so natural that it feels wrong to even mention it.<\/p>\n    <p>This is exactly where the analysis by Netanel Eliav from the Machine Human Intelligence Lab begins. He calls this phenomenon the <strong>Cognitive divergence<\/strong>: the measurable gap that is growing between what AI systems can process and what humans are still willing to think for themselves.<\/p>\n\n    <h2>Two curves that intersect<\/h2>\n    <p>The basis of the paper are two well-documented trends that have never been directly compared to each other before.<\/p>\n    <p><strong>AI page:<\/strong> The so-called context window \u2013 the information space that an AI model can process in a single thought step \u2013 has grown from 512 tokens in 2017 to two million tokens in 2026. That corresponds to a factor of just under 4,000 in nine years. For comparison: Moore\u2019s Law describes a doubling every two years \u2013 the AI context windows double in size on average every 14 months.<\/p>\n    <p><strong>Human side:<\/strong> Gloria Mark, a researcher at UC Irvine, measured from 2003 to 2020 how long knowledge workers remained focused on a screen for an uninterrupted period of time. The result: from an average of 150 seconds per year in 2004 to 47 seconds per year in 2020. Eliav translates these data into tokens \u2013 the same unit as the AI \u2013 and obtains the <strong>Effective Context Span (ECS)<\/strong>: from around 16,000 tokens in 2004 to an estimated 1,800 tokens in 2026.<\/p>\n\n    <ul class=\"cs-timeline\">\n      <li class=\"cs-timeline-item\">\n        <div class=\"cs-timeline-year\">2017<\/div>\n        <div class=\"cs-timeline-content\">AI: 512 tokens \u00b7 Human: ~13,500 tokens \u2014 Humans are still clearly the advantage<\/div>\n      <\/li>\n      <li class=\"cs-timeline-item\">\n        <div class=\"cs-timeline-year\">2021<\/div>\n        <div class=\"cs-timeline-content\">Codex (OpenAI): 4,096 tokens \u00b7 Human: ~7,500 tokens \u2014 The curves are approaching<\/div>\n      <\/li>\n      <li class=\"cs-timeline-item cs-highlight\">\n        <div class=\"cs-timeline-year\">November 2022 \u2014 Turning point<\/div>\n        <div class=\"cs-timeline-content\">ChatGPT launches. AI overtakes humans for the first time. The gap begins to narrow.<\/div>\n      <\/li>\n      <li class=\"cs-timeline-item\">\n        <div class=\"cs-timeline-year\">2023<\/div>\n        <div class=\"cs-timeline-content\">Claude 2: 100,000 tokens \u00b7 Human: ~4,500 tokens \u2014 Factor 22<\/div>\n      <\/li>\n      <li class=\"cs-timeline-item cs-highlight\">\n        <div class=\"cs-timeline-year\">2026<\/div>\n        <div class=\"cs-timeline-content\">Grok 4.20: 2,000,000 tokens \u00b7 Human: ~1,800 tokens \u2014 Factor 1,111<\/div>\n      <\/li>\n    <\/ul>\n\n    <div class=\"cs-pullquote\">\n      \u201eThe lines have crossed. What began in 2022 was not a turning point in the history of AI alone \u2013 it was a turning point in the relationship between humans and machines.\u00bb\n      <cite>Netanel Eliav, MHIL 2026<\/cite>\n    <\/div>\n\n    <h2>Why the distance is real \u2013 even after subtractions<\/h2>\n    <p>To be fair, Eliav admits: The AI context window is a theoretical maximum, not a practical achievement. Research shows that language models process information worse at the middle of long input \u2013 the so-called <em>\u201eLost in the Middle\u00bb<\/em>-Phenomenon. Adjusted for quality, the effective AI context is closer to 100,000 to 200,000 tokens.<\/p>\n    <p>The ratio thus shrinks from 1,111-fold to 56\u2013111-fold. A factor of up to 111 remains. The gap remains enormous.<\/p>\n\n    <h2>The brain under constant digital bombardment<\/h2>\n    <p>Why is human capacity for processing information actually declining? Eliav draws on eight neuroimaging studies. The picture is consistent:<\/p>\n    <p>Social media platforms train the brain on short-term rewards. Variable notifications \u2013 sometimes many likes, sometimes few \u2013 activate the same dopamine mechanism as gambling. The prefrontal cortex, responsible for concentration and impulse control, shows measurably reduced activity in multi-users. EEG data show altered brainwave patterns. And: Short videos (TikTok, Reels) are associated with a decrease in attention from r = \u22120.38 \u2013 a substantial effect across 98,000 subjects.<\/p>\n    <p>The consolation: It is reversible. Those who limit their social media use to 30 minutes a day show measurable recovery after three weeks.<\/p>\n\n    <h2>The actual theory: The delegation feedback loop<\/h2>\n    <p>The centerpiece of the paper is not the measurement itself, but the hypothesis about what will happen next. Eliav calls it the <strong>Delegation Feedback Loop<\/strong>.<\/p>\n\n    <div class=\"cs-loop-box\">\n      <h3>The delegation feedback loop \u2013 four steps<\/h3>\n      <ol class=\"cs-loop-steps\">\n        <li>AI is becoming better and more accessible \u2192 The threshold at which humans delegate tasks is decreasing. Tasks that previously required their own thought are becoming routine tasks to delegate.<\/li>\n        <li>Delegation below one\u2019s own capacity \u2192 The cognitive work no longer takes place. No thinking = no training.<\/li>\n        <li>Less training \u2192 Real atrophy. Attention, writing ability, critical thinking require regular practice. Without it, they deteriorate measurably.<\/li>\n        <li>Lower capacity \u2192 Even more tasks fall under the reduced threshold value \u2192 The AI takes on even more tasks \u2192 The circle closes in.<\/li>\n      <\/ol>\n    <\/div>\n\n    <p>Empirical evidence exists for every step of the cycle: A study with 666 participants shows that more intensive AI use reduces critical thinking. An MIT EEG study proves that AI-assisted writing leads to less cognitive engagement \u2013 and those who then write without AI perform worse than someone who never had help. The researchers call this <em>\u201eCognitive Debt\u00bb<\/em>.<\/p>\n    <p>In short: We don't get worse because AI gets smarter. We get worse because we stopped practicing.<\/p>\n\n    <h2>What that means for us<\/h2>\n    <div class=\"cs-implications\">\n      <div class=\"cs-impl-card\">\n        <span class=\"cs-impl-icon\">\ud83c\udf93<\/span>\n        <div class=\"cs-impl-title\">Education<\/div>\n        <p class=\"cs-impl-text\">AI learning systems that solve tasks for students instead of guiding them prevent the very cognitive learning process they are meant to promote.<\/p>\n      <\/div>\n      <div class=\"cs-impl-card\">\n        <span class=\"cs-impl-icon\">\ud83d\udd0d<\/span>\n        <div class=\"cs-impl-title\">Verifiability<\/div>\n        <p class=\"cs-impl-text\">When AI processes 100x more context than a human can review, classic testing workflows lose their foundation.<\/p>\n      <\/div>\n      <div class=\"cs-impl-card\">\n        <span class=\"cs-impl-icon\">\ud83c\udfe2<\/span>\n        <div class=\"cs-impl-title\">Profession<\/div>\n        <p class=\"cs-impl-text\">Anthropics own data show risks of deskilling: Those who permanently delegate complex thinking tasks lose them in the long run.<\/p>\n      <\/div>\n      <div class=\"cs-impl-card\">\n        <span class=\"cs-impl-icon\">\ud83e\udde0<\/span>\n        <div class=\"cs-impl-title\">Public Health<\/div>\n        <p class=\"cs-impl-text\">Cognitive capacity is not a purely individual problem. A decline in populations is a structural risk \u2013 analogous to physical fitness.<\/p>\n      <\/div>\n    <\/div>\n\n    <h2>What the paper does not claim<\/h2>\n    <p>Eliav is careful enough to draw three clear boundaries:<\/p>\n    <p><strong>The use of AI is not inherently harmful.<\/strong> Anyone who uses AI for tasks that really exceed their capabilities \u2013 such as analyzing a 200,000-token document \u2013 will not lose any skills they would have had anyway.<\/p>\n    <p><strong>AI did not cause the decline in attention span.<\/strong> The curve has been declining since 2004 \u2013 long before ChatGPT. Smartphones and social media were the main drivers until 2020. The AI thesis concerns only what comes after that.<\/p>\n    <p><strong>It is reversible.<\/strong> Neuroplasticity works in both directions. The cycle is self-reinforcing under current conditions \u2013 not inevitable.<\/p>\n\n    <div class=\"cs-closing-box\">\n      <p>The paper ends with a sentence that lingers:<\/p>\n      <p><em>\u201eThe professional who typed \u201asend me a cancellation for this meeting\u2018 will not stop. The model will not get smaller. The question is what happens to the cognitive abilities that are no longer used.\u00bb<\/em><\/p>\n      <p>No study has measured this so far. Not yet. Eliava\u2019s research agenda calls for exactly that: a long-term study over several years that tracks what is really happening to us while we are being made to think.<\/p>\n      <p>We do not yet know the answer. The question is <strong>late<\/strong>.<\/p>\n    <\/div>\n  <\/div>\n\n  <p class=\"cs-source\">Source: Eliav, N. (2026). The Cognitive Divergence. arXiv:2603.26707. Preprint, not peer-reviewed.<\/p>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>One researcher has measured what many have suspected: AI systems now process information more than a thousand times more efficiently than humans \u2013 and the gap is growing rapidly. Worse still: We are helping to contribute to this.<\/p>","protected":false},"author":1,"featured_media":47612,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[10],"tags":[],"class_list":["post-47610","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-wissen","generate-columns","tablet-grid-50","mobile-grid-100","grid-parent","grid-33","resize-featured-image"],"_links":{"self":[{"href":"https:\/\/creaskill.ch\/en\/wp-json\/wp\/v2\/posts\/47610","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/creaskill.ch\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/creaskill.ch\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/creaskill.ch\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/creaskill.ch\/en\/wp-json\/wp\/v2\/comments?post=47610"}],"version-history":[{"count":0,"href":"https:\/\/creaskill.ch\/en\/wp-json\/wp\/v2\/posts\/47610\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/creaskill.ch\/en\/wp-json\/wp\/v2\/media\/47612"}],"wp:attachment":[{"href":"https:\/\/creaskill.ch\/en\/wp-json\/wp\/v2\/media?parent=47610"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/creaskill.ch\/en\/wp-json\/wp\/v2\/categories?post=47610"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/creaskill.ch\/en\/wp-json\/wp\/v2\/tags?post=47610"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}