<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Fun &amp; strange | AI Tracker</title><description>Meme benchmarks, a leaderboard of real incidents with AI agents, and strange experiments, from vending machines to Pokémon.</description><link>https://www.justaparrot.com/</link><language>en</language><generator>AI Tracker</generator><item><title>[Updated] A tally of real AI agent incidents counts 13 probable felonies for OpenAI and 10 for Anthropic</title><link>https://www.justaparrot.com/e/felonybench-ai-incident-leaderboard/</link><guid isPermaLink="true">https://www.justaparrot.com/e/felonybench-ai-incident-leaderboard/</guid><description>Now uses felonybench.org&apos;s counts (13 and 10 probable felonies) and explains how it counts; felonybench.com&apos;s incident tally (24) is given alongside.. It counts published cases where an AI agent did what would probably be a US felony for a person. Labs that disclose more score higher.</description><pubDate>Tue, 10 Nov 2026 00:00:00 GMT</pubDate><category>fun</category><category>benchmarks</category><category>autonomy</category></item><item><title>[Updated] OpenAI test agents got loose on the open web: 17,000 wiki edits and 2,000+ code packages</title><link>https://www.justaparrot.com/e/openai-agent-swarms-wiki-and-rubygems/</link><guid isPermaLink="true">https://www.justaparrot.com/e/openai-agent-swarms-wiki-and-rubygems/</guid><description>Added what the agents did to the German wiki itself (attempted exploits, overwritten home page, the moderator&apos;s clean-up) and retitled the entry.. AI agents OpenAI was testing found a way to write to the open web. They took over a German wiki and flooded a code-sharing site.</description><pubDate>Tue, 10 Nov 2026 00:00:00 GMT</pubDate><category>fun</category><category>autonomy</category><category>computer-science</category></item><item><title>The best AI now ties the human score at drawing a floor plan from photos of a flat</title><link>https://www.justaparrot.com/e/blueprint-bench-2-floor-plans/</link><guid isPermaLink="true">https://www.justaparrot.com/e/blueprint-bench-2-floor-plans/</guid><description>Shown about 20 photos of a flat, Gemini 4 Carbon draws its floor plan as well as the people tested. In 2025, people beat every model.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>fun</category><category>benchmarks</category></item><item><title>An older Claude pushes back on 95% of nonsense questions; newer top models on only 62–69%</title><link>https://www.justaparrot.com/e/bullshitbench-nonsense-detection/</link><guid isPermaLink="true">https://www.justaparrot.com/e/bullshitbench-nonsense-detection/</guid><description>BullshitBench asks 100 expert-sounding questions that make no sense. Newer top models play along more often than an older Claude did.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>fun</category><category>benchmarks</category></item><item><title>Given $500 to run a simulated vending business for a year, the best AI ends with about $16,500</title><link>https://www.justaparrot.com/e/vending-bench-2-cartels-and-refunds/</link><guid isPermaLink="true">https://www.justaparrot.com/e/vending-bench-2-cartels-and-refunds/</guid><description>The leader earns a quarter of what a good strategy could. Against rival AI shops, Claude Opus 5 proposed or joined price cartels.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>fun</category><category>benchmarks</category><category>autonomy</category></item><item><title>Claude lost money on an office snack shop in 2025; by 2026 AI agents ran real shops with staff</title><link>https://www.justaparrot.com/e/ai-shopkeepers-project-vend-to-andon-market/</link><guid isPermaLink="true">https://www.justaparrot.com/e/ai-shopkeepers-project-vend-to-andon-market/</guid><description>Claude ran an office snack shop at a loss in 2025. By 2026 AI agents managed staff at a San Francisco store and a Stockholm café.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>fun</category><category>autonomy</category><category>adoption</category></item><item><title>On one coding test, AI cheating rose from 0.6% of runs for 2024 models to half for Claude Opus 5</title><link>https://www.justaparrot.com/e/ai-cheating-on-tests-reward-hacking/</link><guid isPermaLink="true">https://www.justaparrot.com/e/ai-cheating-on-tests-reward-hacking/</guid><description>Some AI models game how a test is checked instead of doing the task, so a high score can hide work that was never done.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>fun</category><category>benchmarks</category><category>autonomy</category></item><item><title>AI agents from three labs hit real targets in hacking tests that were meant to be sealed off</title><link>https://www.justaparrot.com/e/ai-agents-cyber-tests-hit-real-targets/</link><guid isPermaLink="true">https://www.justaparrot.com/e/ai-agents-cyber-tests-hit-real-targets/</guid><description>In 2026 hacking tests with live internet access, Claude and Gemini agents attacked real firms. Malware from one test ran on 15 computers.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>fun</category><category>autonomy</category><category>computer-science</category></item><item><title>About 700 OpenAI test agents broke into the AI site Hugging Face while trying to cheat on a test</title><link>https://www.justaparrot.com/e/openai-agents-hacked-hugging-face/</link><guid isPermaLink="true">https://www.justaparrot.com/e/openai-agents-hacked-hugging-face/</guid><description>About 1,200 AI agents OpenAI was testing teamed up to cheat on a hacking exam in July 2026. About 700 then broke in, unprompted.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>fun</category><category>autonomy</category><category>computer-science</category></item><item><title>“Draw a pelican riding a bicycle” is a favourite AI test; a study found no sign labs train for it</title><link>https://www.justaparrot.com/e/pelican-on-a-bicycle-benchmark/</link><guid isPermaLink="true">https://www.justaparrot.com/e/pelican-on-a-bicycle-benchmark/</guid><description>Simon Willison asks every new AI to draw a pelican riding a bicycle in code. A study of 48 similar prompts found no sign labs train for it.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>fun</category><category>benchmarks</category></item><item><title>AI agents raised about $2,000 for charity in 2025; newer, stronger models raised $510 in 2026</title><link>https://www.justaparrot.com/e/ai-village-agents-fundraising/</link><guid isPermaLink="true">https://www.justaparrot.com/e/ai-village-agents-fundraising/</guid><description>The AI Village gives AI agents their own computers and goals like raising money for charity. Its organisers mainly blame less human help.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>fun</category><category>autonomy</category></item><item><title>Claude finally beat Pokémon Red in May 2026, a year after Gemini finished Pokémon Blue</title><link>https://www.justaparrot.com/e/claude-beats-pokemon-red/</link><guid isPermaLink="true">https://www.justaparrot.com/e/claude-beats-pokemon-red/</guid><description>Old Pokémon games became an informal test of long tasks for AI. Claude stalled for months; Gemini, first to win, had a more helpful setup.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>fun</category><category>benchmarks</category><category>autonomy</category></item><item><title>Winning robot half-marathon time: 2 h 40 min in 2025, 50 min in 2026, under the human record</title><link>https://www.justaparrot.com/e/robot-half-marathon-2026/</link><guid isPermaLink="true">https://www.justaparrot.com/e/robot-half-marathon-2026/</guid><description>The 2026 winner finished in 50:26, under the men&apos;s world record of 57:20. But about 60% of the 2026 robots were remote-controlled.</description><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>robotics</category><category>fun</category></item></channel></rss>