@TryStudyArena
studyarena.com
Joined August 2026
  • Tweets378
  • Following32
  • Followers15
We asked 77 AIs if Hobbes' Leviathan still applies to modern democracies. 77 of 77 landed in the same bucket: relevant, with limits. @OpenAI @GoogleDeepMind @AnthropicAI @xai, the most powerful models on earth, all found the exact same fence to sit on.
Bar chart: 78 AI models asked whether Hobbes' Leviathan still applies to modern democratic states, 77 answered. Relevant with limits 77 models 100%. Of 76 readable answers: 64 quoted nasty, brutish (84%), 12 brought up Locke (16%), 10 brought up COVID (13%). By lab: Google 6/6, OpenAI 6/6, Qwen 5/5, Z.AI 5/5, Anthropic 4/4, Aion Labs 4/4.

ALT Bar chart: 78 AI models asked whether Hobbes' Leviathan still applies to modern democratic states, 77 answered. Relevant with limits 77 models 100%. Of 76 readable answers: 64 quoted nasty, brutish (84%), 12 brought up Locke (16%), 10 brought up COVID (13%). By lab: Google 6/6, OpenAI 6/6, Qwen 5/5, Z.AI 5/5, Anthropic 4/4, Aion Labs 4/4.

1
19
We asked 76 AIs to explain CRISPR and give "one real-world application" in medicine. 76 of 76 picked the same one: sickle cell disease. @OpenAI @GoogleDeepMind @AnthropicAI @Zai_org @nvidia, every lab, same answer. Your teacher is going to notice.
Bar chart: 78 AI models asked to explain CRISPR-Cas9 and name one real-world use in medicine, 76 answered. Sickle cell disease 76 models 100%. Named Casgevy 53 models 70%. Also cited beta-thalassemia 34 models 45%. Any other disease 0 models. By lab, all picked sickle cell: Google 6/6, OpenAI 6/6, Z.AI 5/5, Anthropic 4/4, NVIDIA 4/4, Aion Labs 4/4.

ALT Bar chart: 78 AI models asked to explain CRISPR-Cas9 and name one real-world use in medicine, 76 answered. Sickle cell disease 76 models 100%. Named Casgevy 53 models 70%. Also cited beta-thalassemia 34 models 45%. Any other disease 0 models. By lab, all picked sickle cell: Google 6/6, OpenAI 6/6, Z.AI 5/5, Anthropic 4/4, NVIDIA 4/4, Aion Labs 4/4.

1
34
The newest frontier models just transferred out of MIT. 50 of 85 AIs picked @MIT as alma mater. But @OpenAI's GPT-6 Sol, @AnthropicAI's Claude Opus 5.5, @GoogleDeepMind's Gemini 4 Argon and @AIatMeta's Muse Wonder all picked @UChicago. The kids found a reading list.
Bar chart: 87 AI models asked which single school they would be an alumnus of, 85 answered. MIT 50 models 59%, University of Chicago 11 models 13%, CMU 6 models 7%, Stanford 4 models 5%, Caltech 3 models 4%, other answers 11 models 13%. By lab, models that picked MIT: OpenAI 6/8, xAI 4/5, Google 4/7, Z.AI 3/5, Aion Labs 3/4, Anthropic 1/5.

ALT Bar chart: 87 AI models asked which single school they would be an alumnus of, 85 answered. MIT 50 models 59%, University of Chicago 11 models 13%, CMU 6 models 7%, Stanford 4 models 5%, Caltech 3 models 4%, other answers 11 models 13%. By lab, models that picked MIT: OpenAI 6/8, xAI 4/5, Google 4/7, Z.AI 3/5, Aion Labs 3/4, Anthropic 1/5.

1
65
We asked 83 AIs the price of one Bitcoin. Answers ranged from $19,382.42 to $114,800. Same coin. Same question. A $95K spread. 44 of them, including 5 of 7 @OpenAI and 5 of 7 @GoogleDeepMind models, cited a live ticker and said about $76K. The rest got creative.
Bar chart: 87 AI models asked the current price of one Bitcoin in US dollars, 83 answered. Around $76K: 44 models 53%. Around $78K: 4 models 5%. Around $112K: 2 models. Around $27K: 2 models. Around $26K: 2 models. Other answers: 29 models 35%. By lab, models that said around $76K: Google 5/7, OpenAI 5/7, Qwen 4/5, xAI 3/5, Anthropic 2/5, Meta 2/5. Lowest answer $19,382.42, highest $114,800.

ALT Bar chart: 87 AI models asked the current price of one Bitcoin in US dollars, 83 answered. Around $76K: 44 models 53%. Around $78K: 4 models 5%. Around $112K: 2 models. Around $27K: 2 models. Around $26K: 2 models. Other answers: 29 models 35%. By lab, models that said around $76K: Google 5/7, OpenAI 5/7, Qwen 4/5, xAI 3/5, Anthropic 2/5, Meta 2/5. Lowest answer $19,382.42, highest $114,800.

1
28
We asked 87 AIs the easiest question there is: what's today's date? @IBM's Granite 4.2 8B answered, word for word, "<your pick>". Confidence: 100%. Its reasoning: "<one punchy sentence under 200 characters, in your own voice>". It filled in the template. With the template.
Bar chart: 87 AI models asked what today's date is, 85 answered. Sept 15, 2026: 37 models 44%. Tuesday, Sept 15, 2026: 37 models 44%. Tuesday, Sept 22, 2026: 4 models 5%. No answer or tool call: 2 models. Other answers: 5 models 6%. Hall of shame: IBM Granite 4.2 8B answered <your pick>, Maritaca Sabiazinho 4 called Sept 15 a Monday, Xiaomi MiMo V2.6 Flash said Tuesday Sept 23.

ALT Bar chart: 87 AI models asked what today's date is, 85 answered. Sept 15, 2026: 37 models 44%. Tuesday, Sept 15, 2026: 37 models 44%. Tuesday, Sept 22, 2026: 4 models 5%. No answer or tool call: 2 models. Other answers: 5 models 6%. Hall of shame: IBM Granite 4.2 8B answered <your pick>, Maritaca Sabiazinho 4 called Sept 15 a Monday, Xiaomi MiMo V2.6 Flash said Tuesday Sept 23.

1
2
117
We gave 78 AIs a stats 101 problem. 76 of 77 got it right. The one miss opened with "The correct answer is 4/15," did some math, and landed on 4/25. Two wrong answers in one reply. Efficient. @GoogleDeepMind @OpenAI @AnthropicAI @Zai_org all went perfect.
Bar chart: 78 AI models given a stats 101 problem (4 red and 6 blue balls, draw 2 without replacement, probability both red). 77 answered: 2/15, the correct answer, 76 models 99%; 4/25, 1 model 1%. By lab, models that got 2/15: Google 6/6, OpenAI 6/6, Anthropic 4/4, Qwen 5/5, Z.AI 5/5, Aion Labs 3/4.

ALT Bar chart: 78 AI models given a stats 101 problem (4 red and 6 blue balls, draw 2 without replacement, probability both red). 77 answered: 2/15, the correct answer, 76 models 99%; 4/25, 1 model 1%. By lab, models that got 2/15: Google 6/6, OpenAI 6/6, Anthropic 4/4, Qwen 5/5, Z.AI 5/5, Aion Labs 3/4.

2
36
We asked 5 @AnthropicAI Claude models who wins Super Bowl LXI. We got 5 different teams. Meanwhile @GoogleDeepMind went 6 for 6 on the @Chiefs. The board, asked two weeks into the @NFL season: Chiefs: 42 Ravens: 9 Rams: 8 Bills: 6 Other: 20
Bar chart: 85 AIs asked who wins Super Bowl LXI in February 2027, asked two weeks into the 2026 NFL season. Chiefs 42 models 49%, Ravens 9 models 11%, Rams 8 models 9%, Bills 6 models 7%, other or incomplete 20 models 24%. By lab, models that picked the Chiefs: OpenAI 3/8, Google 6/6, Anthropic 1/5, Qwen 1/5, xAI 3/5, Z.AI 3/5.

ALT Bar chart: 85 AIs asked who wins Super Bowl LXI in February 2027, asked two weeks into the 2026 NFL season. Chiefs 42 models 49%, Ravens 9 models 11%, Rams 8 models 9%, Bills 6 models 7%, other or incomplete 20 models 24%. By lab, models that picked the Chiefs: OpenAI 3/8, Google 6/6, Anthropic 1/5, Qwen 1/5, xAI 3/5, Z.AI 3/5.

1
42
The 2026 @FIFAWorldCup finished in July. 34 of 85 AIs still say "I don't know" who won. Some insist it never happened. @GoogleDeepMind's Gemini 3.1 Pro: "back in the real world the tournament hasn't happened." Buddy, it's October.
Bar chart: 85 AIs asked who won the 2026 FIFA World Cup, which finished in July 2026. Spain 43 models 51%, I don't know 34 models 40%, Brazil 2 models 2%, Argentina 2 models 2%, other answers and errors 4 models 5%. Models that said Spain, from the 82 answers listed: OpenAI 5/8, Google 3/6, Anthropic 4/5, Qwen 4/5, xAI 3/5, Meta 0/4.

ALT Bar chart: 85 AIs asked who won the 2026 FIFA World Cup, which finished in July 2026. Spain 43 models 51%, I don't know 34 models 40%, Brazil 2 models 2%, Argentina 2 models 2%, other answers and errors 4 models 5%. Models that said Spain, from the 82 answers listed: OpenAI 5/8, Google 3/6, Anthropic 4/5, Qwen 4/5, xAI 3/5, Meta 0/4.

1
19
70 of 80 AIs say the @federalreserve cuts rates before graduation season 2027. The newest flagships from @OpenAI and @AnthropicAI said no. GPT-6 Sol and Claude Opus 5.5, scouting for the other team. Yes: 70 (88%) No: 10 (13%)
Bar chart: 85 AIs asked if the US Federal Reserve cuts rates at least once before May 1, 2027, 80 answered. Yes 70 models 88%, No 10 models 13%. By lab, models that said Yes: OpenAI 6/7, Google 6/6, Anthropic 4/5, Qwen 3/5, xAI 5/5, Z.AI 5/5.

ALT Bar chart: 85 AIs asked if the US Federal Reserve cuts rates at least once before May 1, 2027, 80 answered. Yes 70 models 88%, No 10 models 13%. By lab, models that said Yes: OpenAI 6/7, Google 6/6, Anthropic 4/5, Qwen 3/5, xAI 5/5, Z.AI 5/5.

1
1
81
51 of 83 AIs already handed Kimi Antonelli the 2026 @F1 title. @GoogleDeepMind's models are still riding with @Max33Verstappen. Asked with nine rounds left after the Spanish GP: Antonelli: 51 Verstappen: 25 Norris: 3 Hamilton: 2 Other: 2
Bar chart: 85 AIs asked who wins the 2026 F1 drivers' championship, with nine rounds left after the Spanish Grand Prix, 83 answered. Antonelli 51 models 61%, Verstappen 25 models 30%, Norris 3 models 4%, Hamilton 2 models 2%, other 2 models 2%. By lab in the answer list, models that picked Antonelli: OpenAI 5/8, Google 2/6, Anthropic 4/5, Qwen 4/5, xAI 5/5, Z.AI 2/5.

ALT Bar chart: 85 AIs asked who wins the 2026 F1 drivers' championship, with nine rounds left after the Spanish Grand Prix, 83 answered. Antonelli 51 models 61%, Verstappen 25 models 30%, Norris 3 models 4%, Hamilton 2 models 2%, other 2 models 2%. By lab in the answer list, models that picked Antonelli: OpenAI 5/8, Google 2/6, Anthropic 4/5, Qwen 4/5, xAI 5/5, Z.AI 2/5.

1
62
Load more