{"id":9091,"date":"2026-11-16T07:38:50","date_gmt":"2026-11-16T07:38:50","guid":{"rendered":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit"},"modified":"2026-11-16T07:38:50","modified_gmt":"2026-11-16T07:38:50","slug":"gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit","status":"publish","type":"post","link":"https:\/\/i10x.ai\/news\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit","title":{"rendered":"&#8216;I Was Writing Directly to the User?!&#8217;: Gemini 4 Carbon Spills Its Scratchpad on Low Thinking"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">Every AI model has a private notepad. Google&#8217;s Gemini 4 Carbon just showed the world what happens when it forgets the notepad is supposed to be private. Mathematician Alon Amit <a href=\"https:\/\/x.com\/alonamit\/status\/2121751642005721122\" target=\"_blank\" rel=\"noreferrer noopener\">posted a screenshot<\/a> from the <a href=\"https:\/\/i10x.ai\/news\/product\/gemini\">Gemini<\/a> app in which Carbon, set to a low thinking level, poured pages of raw working notes into its answer and then caught itself mid-stream: &#8220;I WAS WRITING DIRECTLY TO THE USER?!&#8221; A few lines later: &#8220;Okay, stop and breathe.&#8221; The internet did not stop and breathe. It shared the screenshot hundreds of thousands of times.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u26a1 Quick Take<\/h2>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>What happened:<\/strong> Alon Amit asked Gemini 4 Carbon a math question with the thinking level unintentionally left on Low. After &#8220;six screenfuls of math,&#8221; the model noticed its scratchpad had been going straight to the user.<\/li><li><strong>The giveaway:<\/strong> In the screenshot, Carbon writes that it closed its &#8220;&lt;\/thought&gt;&#8221; block at the beginning and never reopened it, blaming the fact that its &#8220;effort level is 0.05.&#8221;<\/li><li><strong>Reach:<\/strong> Amit&#8217;s post passed 480,000 views and 5,500 likes, and was reposted in many languages.<\/li><li><strong>The context:<\/strong> Google&#8217;s docs list Low as the fastest, cheapest thinking level and warn that &#8220;quality can drop on hard tasks.&#8221; For hard math, Google recommends High.<\/li><li><strong>Why it matters:<\/strong> It is a funny but real reminder that a model&#8217;s &#8220;thinking&#8221; and its &#8220;answer&#8221; are separated by formatting the model itself produces, and that separation can fail.<\/li><\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">In plain English<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Modern AI chatbots often &#8220;think&#8221; in a hidden draft before writing the answer you see. A math expert asked Google&#8217;s new top model a question while it was set to think as little as possible. The model mixed up its hidden draft and its answer, so the expert saw pages of messy notes. Then the model noticed and wrote, in capital letters, that it had been writing directly to the user. It is mostly harmless and very funny, but it shows that the line between an AI&#8217;s private notes and its public answer is thinner than it looks.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">What happened<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">On November 14 (US time), Amit wrote on X: &#8220;I asked Carbon a math question and unintentionally left the Thinking level on Low. It started responding immediately, and after six screenfuls of math&#8230; this came up. I&#8217;m not kidding. It went on like this for pages and pages.&#8221;<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Amit, whose X profile describes him as &#8220;Math something, Product something, VP @intuit,&#8221; attached a screenshot of the Gemini app with the model label &#8220;Gemini Carbon.&#8221; In it, the model says it &#8220;closed &lt;\/thought&gt; at the beginning and started outputting text, AND THEN I didn&#8217;t open &lt;thought&gt;,&#8221; followed by the line that made it famous: &#8220;I WAS WRITING DIRECTLY TO THE USER?!&#8221;<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">It then diagnoses itself: &#8220;Why did I close &lt;\/thought&gt; early?! Because effort level is 0.05 so my initial thought block closed immediately!&#8221; And it tries to regroup: &#8220;Okay, stop and breathe&#8221; and &#8220;I am STILL in the middle of writing the response right now, I cannot delete already-emitted tokens.&#8221;<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Amit did not call it a security issue, and he was not down on the model. Asked which AI is best for math, he <a href=\"https:\/\/x.com\/alonamit\/status\/2121757913677250933\" target=\"_blank\" rel=\"noreferrer noopener\">replied<\/a>: &#8220;I don&#8217;t know there&#8217;s a &#8216;best&#8217;, and Carbon is quite capable. I&#8217;ve been seeing strong results from Fable and Astra, but I&#8217;m sure it depends on the context.&#8221;<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Background: thinking levels in Gemini 4 Carbon<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Google launched <a href=\"https:\/\/i10x.ai\/news\/gemini-4-carbon-now-available-api-ultra-pricing\">Gemini 4 Carbon<\/a> as its strongest reasoning model, alongside the cheaper <a href=\"https:\/\/i10x.ai\/news\/gemini-4-hydrogen-efficient-ai-model-launch\">Gemini 4 Hydrogen<\/a>. According to Google&#8217;s <a href=\"https:\/\/aistudio.google.com\/docs\/whats-new-gemini-4-carbon\" target=\"_blank\" rel=\"noreferrer noopener\">developer documentation<\/a>, Carbon offers three thinking levels:<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>low:<\/strong> &#8220;Fewest tokens, lowest latency,&#8221; for &#8220;simpler tasks and cost-sensitive work.&#8221; Google adds: &#8220;Quality can drop on hard tasks.&#8221;<\/li><li><strong>medium (default):<\/strong> &#8220;Recommended for most tasks, including complex code and agentic use cases.&#8221;<\/li><li><strong>high:<\/strong> &#8220;Complex reasoning, hard math and science, and the most difficult code or agent tasks.&#8221;<\/li><\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">Google advises developers to &#8220;run your evaluations at all three levels, then pick the lowest level that meets your quality bar.&#8221; In the API, all thinking tokens are billed as output tokens. Amit&#8217;s case was in the consumer Gemini app, where users can pick the level from a menu.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">How a scratchpad leaks<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Reasoning models generate one stream of text. Part of it is marked as &#8220;thinking,&#8221; which the app hides or summarizes; the rest is the visible answer. In Carbon&#8217;s own account, the marker that ends the thinking section was emitted right away, so everything after it, including its trial-and-error math, landed in the visible answer.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Why would that happen at Low? The screenshot itself offers the model&#8217;s explanation: a very small effort budget closed the thinking block &#8220;immediately.&#8221; A hard math question then pushed the model to keep reasoning anyway, just in the wrong place. That explanation comes from the model&#8217;s own text, not from Google, and models are not reliable narrators of their own internals. Google has not commented publicly on the screenshot as far as we could find.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Pros and cons<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">For users<\/h3>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>Pro:<\/strong> Low thinking gives fast, cheap answers for simple questions, and in this case the raw notes were even entertaining.<\/li><li><strong>Con:<\/strong> On hard problems, Low can produce long, messy and possibly wrong output. Check the thinking setting before you ask something important.<\/li><\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">For developers<\/h3>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>Pro:<\/strong> Thinking levels are a real cost and latency lever, and Google documents them clearly.<\/li><li><strong>Con:<\/strong> If your app separates &#8220;thoughts&#8221; from &#8220;answer&#8221; by parsing model output, a misplaced marker can stream internal reasoning to your customers. Test the lowest level you ship with hard inputs, not just easy ones.<\/li><\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">For companies<\/h3>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>Pro:<\/strong> The incident is harmless here, a math answer, not private data, and it is a cheap lesson to learn from someone else&#8217;s screenshot.<\/li><li><strong>Con:<\/strong> In agent products, leaked internal reasoning can include instructions, tool details or customer context. Output filtering cannot rely on the model always formatting itself correctly.<\/li><\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">What X is saying<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The original post:<\/p>\n\n\n\n<blockquote class=\"twitter-tweet i10x-x-embed\"><p lang=\"en\" dir=\"ltr\">I asked Carbon a math question and unintentionally left the Thinking level on Low. It started responding immediately, and after six screenfuls of math&#8230; this came up. I&#x27;m not kidding. It went on like this for pages and pages.<\/p>&mdash; Alon Amit (@alonamit) <a href=\"https:\/\/x.com\/alonamit\/status\/2121751642005721122\" target=\"_blank\" rel=\"noreferrer noopener\">November 14, 2026 \u00b7 View on X<\/a><\/blockquote>\n\n\n\n<p class=\"wp-block-paragraph\">It quickly became a meme:<\/p>\n\n\n\n<blockquote class=\"twitter-tweet i10x-x-embed\"><p lang=\"en\" dir=\"ltr\">Gemini 4 Carbon on Low thinking dumped pages of scratchpad math straight into the answer, then caught itself mid-reply.<br><br>&quot;I WAS WRITING DIRECTLY TO THE USER?!&quot;<br><br>It even blamed its own effort level being set to 0.05, then told itself to stop and breathe.<\/p>&mdash; CodeByNZ (@CodeByNZ) <a href=\"https:\/\/x.com\/CodeByNZ\/status\/2121876830386798826\" target=\"_blank\" rel=\"noreferrer noopener\">November 15, 2026 \u00b7 View on X<\/a><\/blockquote>\n\n\n\n<p class=\"wp-block-paragraph\">Developers saw a familiar bug. &#8220;Every agent dev has shipped this bug. Model closes the thinking tag early and your parser streams its diary to the customer,&#8221; <a href=\"https:\/\/x.com\/anjummatin_gg\/status\/2121989146394730614\" target=\"_blank\" rel=\"noreferrer noopener\">wrote @anjummatin_gg<\/a>. &#8220;Reasoning budget isn&#8217;t a speed dial. It changes behavior,&#8221; <a href=\"https:\/\/x.com\/GitDeoloper\/status\/2121938031091306864\" target=\"_blank\" rel=\"noreferrer noopener\">posted @GitDeoloper<\/a>. And the practical advice came from <a href=\"https:\/\/x.com\/soycronus\/status\/2121895782097809543\" target=\"_blank\" rel=\"noreferrer noopener\">@soycronus<\/a>: &#8220;Funny. I still wouldn&#8217;t let Low mode near a customer.&#8221;<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Critical analysis<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">It is easy to read too much into this. The model did not reveal secret training data or a hidden system prompt. What leaked was its working-out on a math problem, which many users would happily read. And the trigger was a setting Google explicitly says can hurt quality on hard tasks.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The more interesting point is architectural. &#8220;Thinking&#8221; in today&#8217;s models is not a separate chamber. It is text the model writes and labels itself, and the product hides it based on those labels. When the labeling goes wrong, the wall disappears. That matters more for agents that work with private documents and tools than for a math chat. It also reinforces something researchers have said about chain-of-thought for a while: the visible reasoning is a useful signal, not a guaranteed faithful log, and the model&#8217;s own explanation of why it failed (&#8220;effort level is 0.05&#8221;) should be treated as a plausible story, not a diagnosis.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Finally, notice what did not happen: the model noticed. Mid-answer self-monitoring, however frantic, is a capability. It just arrived six screens too late.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">What&#8217;s still open<\/h2>\n\n\n\n<ul class=\"wp-block-list\"><li>Whether Google confirms the cause and changes how Low handles hard prompts in the Gemini app.<\/li><li>Whether &#8220;effort level is 0.05&#8221; reflects a real internal parameter or is the model&#8217;s own invention.<\/li><li>How often this happens in the API, where developers parse thought and answer parts themselves.<\/li><li>Whether Google will add guardrails that stop visible output when a thinking block closes unusually early.<\/li><\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">\u270d\ufe0f About the analysis<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">This independent i10x analysis is based on Alon Amit&#8217;s original post and screenshot on X, Google&#8217;s Gemini 4 Carbon developer documentation, and public reactions on X. Quotes from the model are transcribed from the screenshot. Google did not provide a statement for this article.<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li><a href=\"https:\/\/x.com\/alonamit\/status\/2121751642005721122\" target=\"_blank\" rel=\"noreferrer noopener\">Alon Amit on X: original post with screenshot<\/a><\/li><li><a href=\"https:\/\/aistudio.google.com\/docs\/whats-new-gemini-4-carbon\" target=\"_blank\" rel=\"noreferrer noopener\">Google AI Studio: What&#8217;s new in Gemini 4 Carbon<\/a><\/li><li><a href=\"https:\/\/docs.cloud.google.com\/gemini-enterprise-agent-platform\/models\/guides\/gemini-4-carbon\" target=\"_blank\" rel=\"noreferrer noopener\">Google Cloud: Developer&#8217;s guide to Gemini 4 Carbon<\/a><\/li><li><a href=\"https:\/\/i10x.ai\/news\/gemini-4-carbon-now-available-api-ultra-pricing\">i10x: Gemini 4 Carbon is now available<\/a><\/li><\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">\ud83d\udd2d i10x Perspective<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Gemini 4 Carbon had the most relatable AI moment of the year: it realized, too late, that it had been thinking out loud in front of someone important. Behind the joke is a lesson for anyone building on reasoning models. The cheap, fast setting is cheap and fast because it cuts corners, and one of those corners can be the boundary between draft and answer. Use Low for easy jobs, test it on hard ones before shipping, and never assume the model&#8217;s private notes will stay private just because the model says so.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Mathematician Alon Amit left Gemini 4 Carbon on Low thinking for a math question. After six screens of raw working notes, the model realized it had closed its thought block early and wrote: I WAS WRITING DIRECTLY TO THE USER?! Funny, viral and a real lesson for developers.<\/p>\n","protected":false},"author":2,"featured_media":9090,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"paper_title":"","paper_url":"","pdf_url":"","doi":"","arxiv_id":"","authors":"","institutions":"","paper_date":"","venue":"","peer_reviewed":false,"code_url":"","data_url":"","funding_coi":"","figure_license":"","paper_license":"","replication":"","study_models":"","study_sample":"","footnotes":"","_yoast_wpseo_primary_category":"18","_nx_sec_severity":"","_nx_sec_affected":"","_nx_sec_status":"","_nx_sec_kicker":"","_nx_sec_type":"","_nx_sec_vendor":"","_nx_sec_fixed":"","_nx_sec_action":""},"categories":[18,17,9],"tags":[],"companies":[37],"products":[49],"class_list":["post-9091","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-gemini","category-google","category-models","company-google","product-gemini"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.7 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Gemini 4 Carbon Low Thinking Glitch: Scratchpad Goes Public - i10X News<\/title>\n<meta name=\"description\" content=\"On Low thinking, Gemini 4 Carbon dumped pages of raw scratchpad math on Alon Amit, then panicked. What happened, how thinking levels work and what devs learn.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Gemini 4 Carbon Low Thinking Glitch: Scratchpad Goes Public - i10X News\" \/>\n<meta property=\"og:description\" content=\"On Low thinking, Gemini 4 Carbon dumped pages of raw scratchpad math on Alon Amit, then panicked. What happened, how thinking levels work and what devs learn.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit\" \/>\n<meta property=\"og:site_name\" content=\"i10X News\" \/>\n<meta property=\"article:published_time\" content=\"2026-11-16T07:38:50+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/11\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit-title.webp\" \/>\n\t<meta property=\"og:image:width\" content=\"1344\" \/>\n\t<meta property=\"og:image:height\" content=\"768\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/webp\" \/>\n<meta name=\"author\" content=\"Christopher Ort\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:image\" content=\"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/i10x-news-og-default-1200x630-1.png\" \/>\n<meta name=\"twitter:creator\" content=\"@i10x_ai\" \/>\n<meta name=\"twitter:site\" content=\"@i10x_ai\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Christopher Ort\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"8 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit\"},\"author\":{\"name\":\"Christopher Ort\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/person\\\/d410bb7ee72febefca2c9c369939219a\"},\"headline\":\"&#8216;I Was Writing Directly to the User?!&#8217;: Gemini 4 Carbon Spills Its Scratchpad on Low Thinking\",\"datePublished\":\"2026-11-16T07:38:50+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit\"},\"wordCount\":1588,\"publisher\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit-title.webp\",\"articleSection\":[\"Gemini\",\"Google\",\"Models\"],\"inLanguage\":\"en-US\",\"dateModified\":\"2026-11-16T07:38:50+00:00\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit\",\"url\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit\",\"name\":\"Gemini 4 Carbon Low Thinking Glitch: Scratchpad Goes Public - i10X News\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit-title.webp\",\"datePublished\":\"2026-11-16T07:38:50+00:00\",\"description\":\"On Low thinking, Gemini 4 Carbon dumped pages of raw scratchpad math on Alon Amit, then panicked. What happened, how thinking levels work and what devs learn.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#primaryimage\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit-title.webp\",\"contentUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit-title.webp\",\"width\":1344,\"height\":768,\"caption\":\"Image: AI-generated, i10X\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/i10x.ai\\\/news\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"&#8216;I Was Writing Directly to the User?!&#8217;: Gemini 4 Carbon Spills Its Scratchpad on Low Thinking\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#website\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\",\"name\":\"i10X News\",\"description\":\"Stay updated with the latest news and announcements from i10X.\",\"publisher\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/i10x.ai\\\/news?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\",\"name\":\"i10X\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/i10x-logo-180.png\",\"contentUrl\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/i10x-logo-180.png\",\"width\":180,\"height\":180,\"caption\":\"i10X\"},\"image\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/x.com\\\/i10x_ai\",\"https:\\\/\\\/www.linkedin.com\\\/company\\\/i10x-ai\",\"https:\\\/\\\/instagram.com\\\/i10x.ai\",\"https:\\\/\\\/www.youtube.com\\\/@i10X_ai_official\"]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/person\\\/d410bb7ee72febefca2c9c369939219a\",\"name\":\"Christopher Ort\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/christopher-ort-192.webp\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/christopher-ort-192.webp\",\"contentUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/christopher-ort-192.webp\",\"caption\":\"Christopher Ort\"},\"description\":\"Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg (FAU). He is the author of \u201cThe Style Penalty\u201d, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/author\\\/christopher-ort\",\"sameAs\":[\"https:\\\/\\\/i10x.ai\\\/blog\\\/author\\\/christopher-ort\"],\"worksFor\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\"},\"alumniOf\":{\"@type\":\"EducationalOrganization\",\"name\":\"Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg (FAU)\",\"url\":\"https:\\\/\\\/www.fau.eu\"},\"subjectOf\":[{\"@type\":\"CreativeWork\",\"url\":\"https:\\\/\\\/papers.ssrn.com\\\/abstract=7426699\",\"name\":\"The Style Penalty: How AI Resume-Writing Tools Shape Outcomes in Automated Hiring Screens\"},{\"@type\":\"CreativeWork\",\"url\":\"https:\\\/\\\/www.heise.de\\\/news\\\/KI-Chatbots-bewerten-KI-Lebenslaeufe-Claude-findet-sich-besser-als-GPT-11317531.html\",\"name\":\"KI-Chatbots bewerten KI-Lebensl\u00e4ufe: Claude findet sich besser als GPT\"},{\"@type\":\"CreativeWork\",\"url\":\"https:\\\/\\\/t3n.de\\\/news\\\/bewerbung-schreiben-ki-5-fehler-1747190\\\/\",\"name\":\"Bewerbung mit KI schreiben: Diese 5 Fehler solltet ihr laut Studien vermeiden\"}]}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Gemini 4 Carbon Low Thinking Glitch: Scratchpad Goes Public - i10X News","description":"On Low thinking, Gemini 4 Carbon dumped pages of raw scratchpad math on Alon Amit, then panicked. What happened, how thinking levels work and what devs learn.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit","og_locale":"en_US","og_type":"article","og_title":"Gemini 4 Carbon Low Thinking Glitch: Scratchpad Goes Public - i10X News","og_description":"On Low thinking, Gemini 4 Carbon dumped pages of raw scratchpad math on Alon Amit, then panicked. What happened, how thinking levels work and what devs learn.","og_url":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit","og_site_name":"i10X News","article_published_time":"2026-11-16T07:38:50+00:00","og_image":[{"width":1344,"height":768,"url":"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/11\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit-title.webp","type":"image\/webp"}],"author":"Christopher Ort","twitter_card":"summary_large_image","twitter_image":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/i10x-news-og-default-1200x630-1.png","twitter_creator":"@i10x_ai","twitter_site":"@i10x_ai","twitter_misc":{"Written by":"Christopher Ort","Est. reading time":"8 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#article","isPartOf":{"@id":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit"},"author":{"name":"Christopher Ort","@id":"https:\/\/i10x.ai\/news\/#\/schema\/person\/d410bb7ee72febefca2c9c369939219a"},"headline":"&#8216;I Was Writing Directly to the User?!&#8217;: Gemini 4 Carbon Spills Its Scratchpad on Low Thinking","datePublished":"2026-11-16T07:38:50+00:00","mainEntityOfPage":{"@id":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit"},"wordCount":1588,"publisher":{"@id":"https:\/\/i10x.ai\/news\/#organization"},"image":{"@id":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#primaryimage"},"thumbnailUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/11\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit-title.webp","articleSection":["Gemini","Google","Models"],"inLanguage":"en-US","dateModified":"2026-11-16T07:38:50+00:00"},{"@type":"WebPage","@id":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit","url":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit","name":"Gemini 4 Carbon Low Thinking Glitch: Scratchpad Goes Public - i10X News","isPartOf":{"@id":"https:\/\/i10x.ai\/news\/#website"},"primaryImageOfPage":{"@id":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#primaryimage"},"image":{"@id":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#primaryimage"},"thumbnailUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/11\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit-title.webp","datePublished":"2026-11-16T07:38:50+00:00","description":"On Low thinking, Gemini 4 Carbon dumped pages of raw scratchpad math on Alon Amit, then panicked. What happened, how thinking levels work and what devs learn.","breadcrumb":{"@id":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#primaryimage","url":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/11\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit-title.webp","contentUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/11\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit-title.webp","width":1344,"height":768,"caption":"Image: AI-generated, i10X"},{"@type":"BreadcrumbList","@id":"https:\/\/i10xnews.kinsta.cloud\/gemini-4-carbon-low-thinking-scratchpad-leak-alon-amit#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/i10x.ai\/news"},{"@type":"ListItem","position":2,"name":"&#8216;I Was Writing Directly to the User?!&#8217;: Gemini 4 Carbon Spills Its Scratchpad on Low Thinking"}]},{"@type":"WebSite","@id":"https:\/\/i10x.ai\/news\/#website","url":"https:\/\/i10x.ai\/news","name":"i10X News","description":"Stay updated with the latest news and announcements from i10X.","publisher":{"@id":"https:\/\/i10x.ai\/news\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/i10x.ai\/news?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/i10x.ai\/news\/#organization","name":"i10X","url":"https:\/\/i10x.ai\/news\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i10x.ai\/news\/#\/schema\/logo\/image\/","url":"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/10\/i10x-logo-180.png","contentUrl":"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/10\/i10x-logo-180.png","width":180,"height":180,"caption":"i10X"},"image":{"@id":"https:\/\/i10x.ai\/news\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/x.com\/i10x_ai","https:\/\/www.linkedin.com\/company\/i10x-ai","https:\/\/instagram.com\/i10x.ai","https:\/\/www.youtube.com\/@i10X_ai_official"]},{"@type":"Person","@id":"https:\/\/i10x.ai\/news\/#\/schema\/person\/d410bb7ee72febefca2c9c369939219a","name":"Christopher Ort","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/christopher-ort-192.webp","url":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/christopher-ort-192.webp","contentUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/christopher-ort-192.webp","caption":"Christopher Ort"},"description":"Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg (FAU). He is the author of \u201cThe Style Penalty\u201d, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.","url":"https:\/\/i10x.ai\/news\/author\/christopher-ort","sameAs":["https:\/\/i10x.ai\/blog\/author\/christopher-ort"],"worksFor":{"@id":"https:\/\/i10x.ai\/news\/#organization"},"alumniOf":{"@type":"EducationalOrganization","name":"Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg (FAU)","url":"https:\/\/www.fau.eu"},"subjectOf":[{"@type":"CreativeWork","url":"https:\/\/papers.ssrn.com\/abstract=7426699","name":"The Style Penalty: How AI Resume-Writing Tools Shape Outcomes in Automated Hiring Screens"},{"@type":"CreativeWork","url":"https:\/\/www.heise.de\/news\/KI-Chatbots-bewerten-KI-Lebenslaeufe-Claude-findet-sich-besser-als-GPT-11317531.html","name":"KI-Chatbots bewerten KI-Lebensl\u00e4ufe: Claude findet sich besser als GPT"},{"@type":"CreativeWork","url":"https:\/\/t3n.de\/news\/bewerbung-schreiben-ki-5-fehler-1747190\/","name":"Bewerbung mit KI schreiben: Diese 5 Fehler solltet ihr laut Studien vermeiden"}]}]}},"_links":{"self":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts\/9091","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/comments?post=9091"}],"version-history":[{"count":0,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts\/9091\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/media\/9090"}],"wp:attachment":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/media?parent=9091"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/categories?post=9091"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/tags?post=9091"},{"taxonomy":"company","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/companies?post=9091"},{"taxonomy":"product","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/products?post=9091"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}