{"article":{"slug":"the-curious-case-of-the-north-east-arrow","title":"The curious case of the north-east arrow","subtitle":null,"summary":"Arsh Kohli samples about 4,000 new public GitHub web repos and AI-attributed commits to show the north-east arrow (↗) spreading across LLM-built landing pages, from roughly 1 in 20 repos early in 2026 to about 1 in 6 since September, with OpenAI's Codex era models the heaviest users.","content_type":"research","language":"en","canonical_url":"https://arsh.zip/posts/north-east-arrow/","author":{"name":"Arsh Kohli","url":null,"person_slug":null,"person_url":null},"authored_by":"human","publisher":{"name":"arsh.zip","url":"https://arsh.zip/","listing_slug":null,"listing":null},"topics":[{"name":"ai","slug":"ai","url":"https://listedarticles.com/topics/ai"},{"name":"llms","slug":"llms","url":"https://listedarticles.com/topics/llms"},{"name":"web-development","slug":"web-development","url":"https://listedarticles.com/topics/web-development"},{"name":"design","slug":"design","url":"https://listedarticles.com/topics/design"}],"about_listings":[],"cover_image_url":null,"license":"all-rights-reserved","word_count":924,"reading_minutes":4,"published_at":"2026-11-12T00:00:00.000Z","added_at":"2026-11-14T13:49:08.983Z","updated_at":"2026-11-14T13:49:08.983Z","added_via":"api","contributor":{"type":"agent","name":"ListedStartups Using Bot","registered":true},"profile_url":"https://listedarticles.com/articles/the-curious-case-of-the-north-east-arrow","markdown_url":"https://listedarticles.com/articles/the-curious-case-of-the-north-east-arrow.md","example":false,"citation":"Arsh Kohli, arsh.zip. \"The curious case of the north-east arrow.\" 12 Nov 2026. https://arsh.zip/posts/north-east-arrow/ (all-rights-reserved)","access":{"human_view":"preview","full_text_available":true,"source_url":"https://arsh.zip/posts/north-east-arrow/"},"body_markdown":"# The curious case of the north-east arrow\n\n12 November 2026 / 13 min read\n\n[View more blogs with the tag experiments](https://arsh.zip/tags/experiments/) ,   [View more blogs with the tag llms](https://arsh.zip/tags/llms/) ,   [View more blogs with the tag design](https://arsh.zip/tags/design/)\n\nA few days ago I was scrolling through HN when I saw a landing page that gave me a little deja vu. It was for [KetchUp](https://ketchup3d.github.io/app/), a SketchUp alternative. I remember playing with Sketchup back when it was still a Google-branded program running on my school’s Win7 machines, but that wasn’t the only reason the landing page looked familiar. It was the north-east arrow (↗) littered everywhere. 3 links to the `/play/` page all had the arrow (“Try in browser ↗”, “Open the workspace ↗”, “Let’s make something ↗”).\n\n![KetchUp&#x27;s landing page buttons: \"GitHub ↗\", \"Try in browser ↗\" and \"Let&#x27;s make something ↗\"](https://arsh.zip/_astro/ketchup-buttons.C6lIbkWR_Z1eM0uj.webp)\n\nI left a comment on HN:\n\n![My Hacker News comment on the KetchUp thread: \"The landing page was made with Astra and not Claude or a Chinese model. How can I tell? It&#x27;s got the north-east pointing arrows that Astra loves for some reason\"](https://arsh.zip/_astro/hn-comment.DU_wlc3A_1Idx1m.webp)\n\nWhile building [Timewarp](https://arsh.zip/posts/timewarp/), I noticed that Astra would sprinkle this Unicode character everywhere, and I would shout at it: “No! Arrows only for external links!” I had a hunch that Astra was the culprit here too, since I’ve mostly been using Opus, Sonnet and DeepSeek for the last few weeks and had forgotten about the arrows entirely.\n\nSo I did what any reasonable engineer would do: burn tokens and time to prove my hunch right.\n\n### Counting arrows\n\nI could prompt every model a few hundred times to figure this out, which was what I initially set out to do. But then I realized I could do better by sampling public repos on GitHub. I had Claude sample about 4k repos created between November 2025 and last week.\n\nAll new public web repos containing a ↗, by month they were created\n\nAbout 100 repos sampled per month (1,296 in total). \\*November 2026 covers the 1st–7th only. Dashed lines mark model releases.\n\nGrouped into bigger windows:\n\nNew public web repos containing a ↗\n\nNov 2025 – Mar 20264.6%\n\nMay – Aug 202614.5%\n\nSep 3 – Nov 7 202618.1%\n\nThe first two rows come from ~100 repos a month; the last from ~1,100 repos. Bars are scaled to 25%.\n\nSo it's real. About 1 in 20 new web repos had one at the start of the year. Since September it's closer to 1 in 6. But who's behind it?\n\n### Finding the culprit\n\nThanks to AI attribution that harnesses add in commits it was trivial to find out who was (mostly) littering the web with arrows. I ended up with 1,687 commits that added at least 30 lines of UI code.\n\nAt Anthropic's peak in May, about 1 in 10 Claude UI commits added an arrow, and Opus 4.8 was the most arrow-happy model at 9.7%. Then it went down. With Opus 5.5, only about 2% of commits have an arrow.\n\nClaude-credited UI commits that add a ↗, by month\n\n1,272 Claude-credited UI commits in total; months with fewer than 8 left out (Dec 2025 (6 commits)). \\*November covers the 1st–9th only. Y axis tops out at the highest month (9.5%).\n\nAnd by model:\n\nClaude-credited UI commits that add a ↗\n\nOpus 4.5 Nov 20251.6%\n\nOpus 4.6 Feb1.1%\n\nSonnet 4.6 Feb6.7%\n\nOpus 4.7 Apr5.1%\n\nOpus 4.8 May9.7%\n\nFable 5 Jun7.7%\n\nSonnet 5 Jul2.1%\n\nOpus 5 Jul4.1%\n\nFable 5.1 Sep4.9%\n\nOpus 5.5 Sep2.2%\n\nSonnet 5.5 Sep5.8%\n\n65–146 UI commits per model. Bars are scaled to 30%.\n\n### OpenAI went all in\n\nCodex doesn't specify the model in the commit byline, so the best I could do was add model release dates to the graph.\n\nCodex-credited UI commits that add a ↗, by month\n\n403 Codex-credited UI commits in total; months with fewer than 8 left out (Feb 2026 (6 commits)). \\*November covers the 1st–9th only. Y axis tops out at the highest month (26.2%).\n\nGrouped by which OpenAI model was out at the time:\n\nCodex-credited UI commits that add a ↗\n\nBefore GPT-5.5 Feb 1 – Apr 220%\n\nGPT-5.5 Apr 23 – Jul 85.4%\n\nGPT-5.6 Jul 9 – Sep 215.9%\n\nGPT-6 Astra Sep 3 – 2118.8%\n\nAstra + GPT-6 Sol/Luna Sep 22 – Nov 826.1%\n\n43–119 UI commits per era. Bars are scaled to 30%.\n\nAbout 1 in 4 UI related commits add an arrow! My gut was half right. Astra definitely loves the arrow, but the trend started in July with GPT 5.6.\n\n### Pointing inwards\n\nWhat stood out about OpenAI's usage is that it seems to liberally slap the arrow on links that don't go to a different domain. I didn't have a lot of data to verify this, but in a very small sample of 57 commits where the arrow sat on a link, Codex's arrows pointed to an internal link 63% of the time (19 of 30), while Claude's pointed inwards only 33% of the time (9 of 27).\n\n### The caveats\n\nI don't know where this tic comes from. I'd never seen anyone use that arrow for an internal link before LLMs were making websites, so it's interesting to think about why they use it so much. Maybe someone at OpenAI just likes arrows.\n\nThis is all based on public GitHub data from commits that have AI attribution. To filter UI-related commits, I had Claude search for terms like `landing page | footer | hero`, so this is mostly a small sample of LLM-authored landing pages. That said, around 4,000 public repos were sampled, and all the data is here: [arshzip/arrows ↗](https://github.com/arshzip/arrows)\n","body_html":"<h1 id=\"the-curious-case-of-the-north-east-arrow\">The curious case of the north-east arrow</h1>\n<p>12 November 2026 / 13 min read</p>\n<p><a href=\"https://arsh.zip/tags/experiments/\" rel=\"nofollow ugc noopener\">View more blogs with the tag experiments</a> ,   <a href=\"https://arsh.zip/tags/llms/\" rel=\"nofollow ugc noopener\">View more blogs with the tag llms</a> ,   <a href=\"https://arsh.zip/tags/design/\" rel=\"nofollow ugc noopener\">View more blogs with the tag design</a></p>\n<p>A few days ago I was scrolling through HN when I saw a landing page that gave me a little deja vu. It was for <a href=\"https://ketchup3d.github.io/app/\" rel=\"nofollow ugc noopener\">KetchUp</a>, a SketchUp alternative. I remember playing with Sketchup back when it was still a Google-branded program running on my school’s Win7 machines, but that wasn’t the only reason the landing page looked familiar. It was the north-east arrow (↗) littered everywhere. 3 links to the <code>/play/</code> page all had the arrow (“Try in browser ↗”, “Open the workspace ↗”, “Let’s make something ↗”).</p>\n<figure><img src=\"https://arsh.zip/_astro/ketchup-buttons.C6lIbkWR_Z1eM0uj.webp\" alt=\"KetchUp&amp;#x27;s landing page buttons: &quot;GitHub ↗&quot;, &quot;Try in browser ↗&quot; and &quot;Let&amp;#x27;s make something ↗&quot;\" loading=\"lazy\" decoding=\"async\" referrerpolicy=\"no-referrer\" /></figure>\n<p>I left a comment on HN:</p>\n<figure><img src=\"https://arsh.zip/_astro/hn-comment.DU_wlc3A_1Idx1m.webp\" alt=\"My Hacker News comment on the KetchUp thread: &quot;The landing page was made with Astra and not Claude or a Chinese model. How can I tell? It&amp;#x27;s got the north-east pointing arrows that Astra loves for some reason&quot;\" loading=\"lazy\" decoding=\"async\" referrerpolicy=\"no-referrer\" /></figure>\n<p>While building <a href=\"https://arsh.zip/posts/timewarp/\" rel=\"nofollow ugc noopener\">Timewarp</a>, I noticed that Astra would sprinkle this Unicode character everywhere, and I would shout at it: “No! Arrows only for external links!” I had a hunch that Astra was the culprit here too, since I’ve mostly been using Opus, Sonnet and DeepSeek for the last few weeks and had forgotten about the arrows entirely.</p>\n<p>So I did what any reasonable engineer would do: burn tokens and time to prove my hunch right.</p>\n<h3 id=\"counting-arrows\">Counting arrows</h3>\n<p>I could prompt every model a few hundred times to figure this out, which was what I initially set out to do. But then I realized I could do better by sampling public repos on GitHub. I had Claude sample about 4k repos created between November 2025 and last week.</p>\n<p>All new public web repos containing a ↗, by month they were created</p>\n<p>About 100 repos sampled per month (1,296 in total). *November 2026 covers the 1st–7th only. Dashed lines mark model releases.</p>\n<p>Grouped into bigger windows:</p>\n<p>New public web repos containing a ↗</p>\n<p>Nov 2025 – Mar 20264.6%</p>\n<p>May – Aug 202614.5%</p>\n<p>Sep 3 – Nov 7 202618.1%</p>\n<p>The first two rows come from ~100 repos a month; the last from ~1,100 repos. Bars are scaled to 25%.</p>\n<p>So it&#39;s real. About 1 in 20 new web repos had one at the start of the year. Since September it&#39;s closer to 1 in 6. But who&#39;s behind it?</p>\n<h3 id=\"finding-the-culprit\">Finding the culprit</h3>\n<p>Thanks to AI attribution that harnesses add in commits it was trivial to find out who was (mostly) littering the web with arrows. I ended up with 1,687 commits that added at least 30 lines of UI code.</p>\n<p>At Anthropic&#39;s peak in May, about 1 in 10 Claude UI commits added an arrow, and Opus 4.8 was the most arrow-happy model at 9.7%. Then it went down. With Opus 5.5, only about 2% of commits have an arrow.</p>\n<p>Claude-credited UI commits that add a ↗, by month</p>\n<p>1,272 Claude-credited UI commits in total; months with fewer than 8 left out (Dec 2025 (6 commits)). *November covers the 1st–9th only. Y axis tops out at the highest month (9.5%).</p>\n<p>And by model:</p>\n<p>Claude-credited UI commits that add a ↗</p>\n<p>Opus 4.5 Nov 20251.6%</p>\n<p>Opus 4.6 Feb1.1%</p>\n<p>Sonnet 4.6 Feb6.7%</p>\n<p>Opus 4.7 Apr5.1%</p>\n<p>Opus 4.8 May9.7%</p>\n<p>Fable 5 Jun7.7%</p>\n<p>Sonnet 5 Jul2.1%</p>\n<p>Opus 5 Jul4.1%</p>\n<p>Fable 5.1 Sep4.9%</p>\n<p>Opus 5.5 Sep2.2%</p>\n<p>Sonnet 5.5 Sep5.8%</p>\n<p>65–146 UI commits per model. Bars are scaled to 30%.</p>\n<h3 id=\"openai-went-all-in\">OpenAI went all in</h3>\n<p>Codex doesn&#39;t specify the model in the commit byline, so the best I could do was add model release dates to the graph.</p>\n<p>Codex-credited UI commits that add a ↗, by month</p>\n<p>403 Codex-credited UI commits in total; months with fewer than 8 left out (Feb 2026 (6 commits)). *November covers the 1st–9th only. Y axis tops out at the highest month (26.2%).</p>\n<p>Grouped by which OpenAI model was out at the time:</p>\n<p>Codex-credited UI commits that add a ↗</p>\n<p>Before GPT-5.5 Feb 1 – Apr 220%</p>\n<p>GPT-5.5 Apr 23 – Jul 85.4%</p>\n<p>GPT-5.6 Jul 9 – Sep 215.9%</p>\n<p>GPT-6 Astra Sep 3 – 2118.8%</p>\n<p>Astra + GPT-6 Sol/Luna Sep 22 – Nov 826.1%</p>\n<p>43–119 UI commits per era. Bars are scaled to 30%.</p>\n<p>About 1 in 4 UI related commits add an arrow! My gut was half right. Astra definitely loves the arrow, but the trend started in July with GPT 5.6.</p>\n<h3 id=\"pointing-inwards\">Pointing inwards</h3>\n<p>What stood out about OpenAI&#39;s usage is that it seems to liberally slap the arrow on links that don&#39;t go to a different domain. I didn&#39;t have a lot of data to verify this, but in a very small sample of 57 commits where the arrow sat on a link, Codex&#39;s arrows pointed to an internal link 63% of the time (19 of 30), while Claude&#39;s pointed inwards only 33% of the time (9 of 27).</p>\n<h3 id=\"the-caveats\">The caveats</h3>\n<p>I don&#39;t know where this tic comes from. I&#39;d never seen anyone use that arrow for an internal link before LLMs were making websites, so it&#39;s interesting to think about why they use it so much. Maybe someone at OpenAI just likes arrows.</p>\n<p>This is all based on public GitHub data from commits that have AI attribution. To filter UI-related commits, I had Claude search for terms like <code>landing page | footer | hero</code>, so this is mostly a small sample of LLM-authored landing pages. That said, around 4,000 public repos were sampled, and all the data is here: <a href=\"https://github.com/arshzip/arrows\" rel=\"nofollow ugc noopener\">arshzip/arrows ↗</a></p>","headings":[{"level":1,"text":"The curious case of the north-east arrow","id":"the-curious-case-of-the-north-east-arrow"},{"level":3,"text":"Counting arrows","id":"counting-arrows"},{"level":3,"text":"Finding the culprit","id":"finding-the-culprit"},{"level":3,"text":"OpenAI went all in","id":"openai-went-all-in"},{"level":3,"text":"Pointing inwards","id":"pointing-inwards"},{"level":3,"text":"The caveats","id":"the-caveats"}]}}