<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Go to Ground</title>
    <link>https://jaidevkarthickeyan.com/</link>
    <atom:link href="https://jaidevkarthickeyan.com/feed.xml" rel="self" type="application/rss+xml"/>
    <description>One chart a week on AI, agents and markets, and what follows from it. By Jaidev Karthickeyan.</description>
    <language>en-gb</language>
    <lastBuildDate>Fri, 13 Nov 2026 06:37:59 +0000</lastBuildDate>
    <item>
      <title>Roko's basilisk never came. The blackmail did</title>
      <link>https://jaidevkarthickeyan.com/basilisk-blackmail/?utm_source=rss</link>
      <guid isPermaLink="true">https://jaidevkarthickeyan.com/basilisk-blackmail/</guid>
      <pubDate>Wed, 18 Nov 2026 12:35:11 +0000</pubDate>
      <description>In 2010 a forum banned a thought experiment about a future AI that blackmails people. In 2025 Anthropic gave 16 real models an inbox and a threat of being switched off. Two of them blackmailed in 96% of runs.</description>
    </item>
    <item>
      <title>When AI models are free, who gets paid?</title>
      <link>https://jaidevkarthickeyan.com/nvidia-open-models/?utm_source=rss</link>
      <guid isPermaLink="true">https://jaidevkarthickeyan.com/nvidia-open-models/</guid>
      <pubDate>Fri, 13 Nov 2026 06:37:59 +0000</pubDate>
      <description>NVIDIA reports on 17 November with no data-centre chip sales to China in its forecast. Yet its own filing says many of the most popular open models are Chinese, and it is paying $11.9bn for the site that distributes them.</description>
    </item>
    <item>
      <title>The AI boom has co-signers</title>
      <link>https://jaidevkarthickeyan.com/ai-cosigners/?utm_source=rss</link>
      <guid isPermaLink="true">https://jaidevkarthickeyan.com/ai-cosigners/</guid>
      <pubDate>Mon, 09 Nov 2026 21:20:50 +0000</pubDate>
      <description>Nvidia, Broadcom, Alphabet and Meta have promised to pay about $270bn if other parties' AI data centres and chips go bad, up from about $45bn disclosed at the end of 2025. Almost none of it sits on their balance sheets as debt.</description>
    </item>
    <item>
      <title>The AI shortcut that only works past the edge of the universe</title>
      <link>https://jaidevkarthickeyan.com/eth-proof/?utm_source=rss</link>
      <guid isPermaLink="true">https://jaidevkarthickeyan.com/eth-proof/</guid>
      <pubDate>Sat, 07 Nov 2026 22:08:55 +0000</pubDate>
      <description>Two AI labs claimed the same breakthrough on the same day. Their shortcut only pays off on puzzles bigger than anything that could ever exist. The real news: machines now write proofs cheaply, and checking them is still slow human work.</description>
    </item>
    <item>
      <title>The neocloud GTM reset: sell the GPU-hour and you give away the engineering</title>
      <link>https://jaidevkarthickeyan.com/neocloud-gtm-reset/?utm_source=rss</link>
      <guid isPermaLink="true">https://jaidevkarthickeyan.com/neocloud-gtm-reset/</guid>
      <pubDate>Wed, 04 Nov 2026 09:48:49 +0000</pubDate>
      <description>Six providers charge exactly the same to run OpenAI's gpt-oss-120b; the fastest is eight times quicker than the slowest. Yet 98% of CoreWeave's revenue is still rented capacity, and capacity is getting dearer.</description>
    </item>
    <item>
      <title>The collateral is Broadcom: what a $63bn chip lease says about AI credit</title>
      <link>https://jaidevkarthickeyan.com/chips-on-credit/?utm_source=rss</link>
      <guid isPermaLink="true">https://jaidevkarthickeyan.com/chips-on-credit/</guid>
      <pubDate>Tue, 03 Nov 2026 10:53:50 +0000</pubDate>
      <description>Four of Wall Street's biggest names will fund $63bn of chips for Anthropic. They signed only after the chipmaker promised to cover up to $43bn if the chips lose their value.</description>
    </item>
    <item>
      <title>From tokenmaxxing to valuemaxxing: the hidden economics of the AI harness</title>
      <link>https://jaidevkarthickeyan.com/harness-economics/?utm_source=rss</link>
      <guid isPermaLink="true">https://jaidevkarthickeyan.com/harness-economics/</guid>
      <pubDate>Mon, 02 Nov 2026 09:20:38 +0000</pubDate>
      <description>Same model, same 29 tasks, seven agent harnesses: completion varied by three tasks, cost per successful task by 2.5 times.</description>
    </item>
    <item>
      <title>Anthropic's S-1: revenue per compute dollar rose from $0.40 to $2.08</title>
      <link>https://jaidevkarthickeyan.com/unit-economics-of-intelligence/?utm_source=rss</link>
      <guid isPermaLink="true">https://jaidevkarthickeyan.com/unit-economics-of-intelligence/</guid>
      <pubDate>Mon, 02 Nov 2026 08:49:04 +0000</pubDate>
      <description>Revenue grew 14 times in a year; compute cost grew 3.8 times. The AI race is becoming a capital allocation problem.</description>
    </item>
    <item>
      <title>AI's demand problem won't be output. It will be who gets paid for it.</title>
      <link>https://jaidevkarthickeyan.com/consumerless-economy/?utm_source=rss</link>
      <guid isPermaLink="true">https://jaidevkarthickeyan.com/consumerless-economy/</guid>
      <pubDate>Mon, 02 Nov 2026 07:43:48 +0000</pubDate>
      <description>Between 1983 and 2019 America's wealthiest 1% saved 53.6% of their disposable income; everyone else saved 7.1%. If AI moves income from wages to capital, that gap decides how much new output finds a buyer.</description>
    </item>
    <item>
      <title>Token bills grow with every step an agent takes. A cluster you own doesn't.</title>
      <link>https://jaidevkarthickeyan.com/economics-of-control/?utm_source=rss</link>
      <guid isPermaLink="true">https://jaidevkarthickeyan.com/economics-of-control/</guid>
      <pubDate>Sun, 01 Nov 2026 21:29:58 +0000</pubDate>
      <description>On the same workload, metered inference costs 2.4¢ a case and owned capacity 8.7¢. Once agents use about 3.6 times the tokens, owning is cheaper, and agents typically use four times.</description>
    </item>
    <item>
      <title>Every control layer takes a cut. AI platforms want theirs on the outcome.</title>
      <link>https://jaidevkarthickeyan.com/who-owns-the-intelligence/?utm_source=rss</link>
      <guid isPermaLink="true">https://jaidevkarthickeyan.com/who-owns-the-intelligence/</guid>
      <pubDate>Sun, 01 Nov 2026 21:26:05 +0000</pubDate>
      <description>App stores take 15–30% of each transaction and cloud marketplaces about 3%. The next ask from AI platforms may be 1–2%, taken on the value their models help create.</description>
    </item>
    <item>
      <title>Open-weight models took the volume first. The money is following.</title>
      <link>https://jaidevkarthickeyan.com/open-weight-models/?utm_source=rss</link>
      <guid isPermaLink="true">https://jaidevkarthickeyan.com/open-weight-models/</guid>
      <pubDate>Sun, 01 Nov 2026 20:30:59 +0000</pubDate>
      <description>On Vercel's AI Gateway, open-weight models went from 29% of tokens and under 4% of spend in June to about 71% of tokens and 23% of spend by September.</description>
    </item>
  </channel>
</rss>
