{
 "about": "Reviewer takes on AI models released 20 October to 19 November 2026, scored for stance, and how benchmark rankings compare with them. Data behind springprompt.com/blog/benchmarks-vs-vibes.",
 "collected": "2026-11-19",
 "scale": "Stance from -2 (strongly negative) to +2 (strongly positive), judged from the quoted words only.",
 "rated_by": "Both raters were claude-opus-5-5 (Anthropic). The first scored each quote while collecting it; the second scored the quotes alone, blind to the first rater's scores and to all benchmark figures.",
 "rater_agreement": {
  "n": 287,
  "spearman": 0.93,
  "exact": 0.791,
  "within1": 0.997,
  "sign_flip": 0,
  "basis_agree": 0.958
 },
 "notes": "Quotes are verbatim. Video quotes come from auto-generated captions and keep their spellings and a leading [h:mm:ss] timestamp. Long-form quotes (id L...) were checked by script against the transcript or page; X posts (id X...) were copied from the opened post. basis: only hands-on records count towards the vibe score. Third-party benchmark values are not republished here; only the rank correlations and head-to-head agreement we computed from them.",
 "reactions": [
  {
   "id": "L000",
   "voice": "Theo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=9KvI-zg4mfo&t=2084s",
   "quote": "[0:34:44] I do not like 62 Astra. I would not choose this over Opus 55 for any of my work as a dev.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -1
  },
  {
   "id": "L001",
   "voice": "Theo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=9KvI-zg4mfo&t=2287s",
   "quote": "[0:38:07] I've already built the habit of switching to opus 55 at the end of a long 62 Astro run just to get a readable summary of what it said.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -1
  },
  {
   "id": "L002",
   "voice": "Theo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=9KvI-zg4mfo&t=1011s",
   "quote": "[0:16:51] And I will say computer use of this model absolutely cracked. It flies.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L003",
   "voice": "Theo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=9KvI-zg4mfo&t=1676s",
   "quote": "[0:27:56] The text this model wrote, despite getting the specific instruction to analyze all of the dialogue in the game and try to match it in terms of tone and character like actions, it just failed outright.",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -1
  },
  {
   "id": "L004",
   "voice": "Theo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=9KvI-zg4mfo&t=756s",
   "quote": "[0:12:36] it's still just bad in comparison to like Fable 5 2 with like the nice animations for these types of things. Like this is significantly better than anything we've seen OpenAI models make ever.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L005",
   "voice": "Theo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=9KvI-zg4mfo&t=23s",
   "quote": "[0:00:23] In my own personal benches, 62 Astra shows dominance across the board, and 62 Souls a very close second place",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L006",
   "voice": "Theo",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=9KvI-zg4mfo&t=23s",
   "quote": "[0:00:23] In my own personal benches, 62 Astra shows dominance across the board, and 62 Souls a very close second place",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L007",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=9KvI-zg4mfo&t=1881s",
   "quote": "[0:31:21] Significantly better water, significantly better fitting models. The planet size and the just overall experience of everything is significantly better.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6.2 Astra",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L008",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=yh5IAUow4Hc&t=571s",
   "quote": "[0:09:31] And I found that Fable 5 2 is actually underperforming Opus and Sonnet and massively outperforming OpenAI models in this particular eval set.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    },
    {
     "model": "Claude Sonnet 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L009",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=yh5IAUow4Hc&t=2000s",
   "quote": "[0:33:20] But I have found I've had to be in the loop a lot more with this model. … It seems like this model almost needs it a bit more than Opus did.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L010",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=yh5IAUow4Hc&t=1954s",
   "quote": "[0:32:34] It has its moments, but yeah, this model really needs the design skill or you're going to end up with fable slop.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": -1
  },
  {
   "id": "L011",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=yh5IAUow4Hc&t=2066s",
   "quote": "[0:34:26] It also seems better at writing prose in general and talking to models",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L012",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=yh5IAUow4Hc&t=2179s",
   "quote": "[0:36:19] What I'm trying to say is this model is pleasant to work with. It does what you tell it and it does it relatively well.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L013",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=yh5IAUow4Hc&t=1863s",
   "quote": "[0:31:03] It writes way faster too, as I mentioned. The TPS is way better. I'm impressed.",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L014",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-13",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=XtykeVoSmP8&t=865s",
   "quote": "[0:14:25] There's a lot of work that Opus is better than Fable 4 for that reason. It will find the things that Fable would have glossed over.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": -1
  },
  {
   "id": "L015",
   "voice": "Theo",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-11",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=OBNVzWYMKbE&t=1549s",
   "quote": "[0:25:49] Would I ever pick this model as my default for coding? No. Absolutely not. It's just Opus is unbeatable still.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    },
    {
     "model": "Gemini 4 Argon",
     "direction": "better"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L016",
   "voice": "Theo",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-11",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=OBNVzWYMKbE&t=356s",
   "quote": "[0:05:56] This model is slow and it sucks to use cuz it's slow.",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -1
  },
  {
   "id": "L017",
   "voice": "Theo",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-11",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=OBNVzWYMKbE&t=1708s",
   "quote": "[0:28:28] The more you look, the worse it gets. Oh god. Yeah. Uh don't use this model for 3D.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "L018",
   "voice": "Theo",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-11",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=OBNVzWYMKbE&t=1437s",
   "quote": "[0:23:57] So, uh, yeah, this model might actually be a pretty good tool for incident response in security stuff.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L019",
   "voice": "Theo",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-13",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=5BrgFY5jIn4&t=935s",
   "quote": "[0:15:35] But right now the model's not good enough to justify using. And even if it was, the cost isn't good enough to justify spending. And even if it was, the speed is so slow that you'll be miserable using it.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "L020",
   "voice": "Bijan Bowen",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=0Lf_uKY9zvw&t=2284s",
   "quote": "[0:38:04] The frame rate is is bad, and this is on the lowest graphics setting, but the way that it took the source material and replicated it in 3js is fantastic.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L021",
   "voice": "Bijan Bowen",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=0Lf_uKY9zvw&t=2425s",
   "quote": "[0:40:25] This thing is very very token heavy and it does not necessarily seem to benefit from the usage sipping that Opus 55 or Sonnet 55 did.",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    },
    {
     "model": "Claude Sonnet 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L022",
   "voice": "Bijan Bowen",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=0GSvcg9wncc&t=2222s",
   "quote": "[0:37:02] I'm extremely impressed with this model, especially considering at least till November 30th, it is a quarter of the price of Fable.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Gemini 4 Argon",
     "direction": "better"
    },
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L023",
   "voice": "Bijan Bowen",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=0GSvcg9wncc&t=861s",
   "quote": "[0:14:21] that is to date absolutely hands down the best replication of the source image that we've seen.",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L024",
   "voice": "Bijan Bowen",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=jm_h5nRZoCg&t=2083s",
   "quote": "[0:34:43] I like it. I think it's excellent",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L025",
   "voice": "Bijan Bowen",
   "model": "Qwen4-Max",
   "date": "2026-11-16",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=atz4gzB7ykY&t=1724s",
   "quote": "[0:28:44] I just have to be honest in that I don't necessarily see this stacking up with the models that it's being benchmarked against in the benchmark charts. Maybe Kimi K3.1 Pro is more of a fair comparison, but not Opus 55",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    },
    {
     "model": "Kimi K3.1 Pro",
     "direction": "same"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L026",
   "voice": "Bijan Bowen",
   "model": "Qwen4-Max",
   "date": "2026-11-16",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=atz4gzB7ykY&t=619s",
   "quote": "[0:10:19] So overall, this was a very, very, very surprising and impressive result with the robot arm. Basically one of the best ever.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L027",
   "voice": "Bijan Bowen",
   "model": "Qwen4-Max",
   "date": "2026-11-16",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=atz4gzB7ykY&t=934s",
   "quote": "[0:15:34] Visually definitely not at a level of what a frontier model I would expect at this point. Though nor was the Fable 5 2 one.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "same"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L028",
   "voice": "Bijan Bowen",
   "model": "Qwen4-Max",
   "date": "2026-11-16",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=atz4gzB7ykY&t=1770s",
   "quote": "[0:29:30] And I have to say at the API pricing, it has stiff competition between like GPT-61 Terra as well as Gemini Hydrogen.",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6.1 Terra",
     "direction": "worse"
    },
    {
     "model": "Gemini 4 Hydrogen",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": 0
  },
  {
   "id": "L029",
   "voice": "Bijan Bowen",
   "model": "Gemini 4 Hydrogen",
   "date": "2026-11-13",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=prA0iTPDqfo&t=1830s",
   "quote": "[0:30:30] So I just don't exactly know why this model so heavily destroyed every other model that we've had try to replicate this in 3D.",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L030",
   "voice": "Bijan Bowen",
   "model": "Gemini 4 Hydrogen",
   "date": "2026-11-13",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=prA0iTPDqfo&t=638s",
   "quote": "[0:10:38] overall acceptable for a flash model, but not really at the level of polish that I'd hoped to see.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Sonnet 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L031",
   "voice": "Bijan Bowen",
   "model": "Gemini 4 Hydrogen",
   "date": "2026-11-13",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=prA0iTPDqfo&t=1876s",
   "quote": "[0:31:16] So the entirety of everything we saw today took 4 % of the weekly usage, which I think is really darn good.",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L032",
   "voice": "Bijan Bowen",
   "model": "Hy4",
   "date": "2026-11-06",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=GBjDU9Ct0VQ&t=1933s",
   "quote": "[0:32:13] I'm just I'm blown away by this.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L033",
   "voice": "Bijan Bowen",
   "model": "Hy4",
   "date": "2026-11-06",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=GBjDU9Ct0VQ&t=1848s",
   "quote": "[0:30:48] This is a solid model. It seems to be pretty darn good at C++.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L034",
   "voice": "Bijan Bowen",
   "model": "GPT-6.1 Terra",
   "date": "2026-10-27",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=VHJfY-c7VkA&t=1721s",
   "quote": "[0:28:41] So, I'm quite happy to report that I like this model.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6 Astra",
     "direction": "same"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L035",
   "voice": "Bijan Bowen",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-24",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=4MRMLerpcQc&t=2596s",
   "quote": "[0:43:16] Overall, that was acceptable. It was nothing that blew me away. … what I'm trying to say is this was not at the level of Fable 5 or GPT-6 Astra.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "GPT-6 Astra",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": 0
  },
  {
   "id": "L036",
   "voice": "Bijan Bowen",
   "model": "DeepSeek Flash 1017",
   "date": "2026-10-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=dk1OY5z08_Y&t=2050s",
   "quote": "[0:34:10] In lieu of a traditional results overview, this is a significant improvement in its 3D capabilities based off of what I'm seeing right here.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "L037",
   "voice": "Every",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=GHSXnRIbLw4&t=68s",
   "quote": "[0:01:08] Overall, this is absolute S tier.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L038",
   "voice": "Every",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=GHSXnRIbLw4&t=91s",
   "quote": "[0:01:31] Coding, a S tier, like S plus tier model. It actually changes the game for what's possible in software engineering.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L039",
   "voice": "Every",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=GHSXnRIbLw4&t=91s",
   "quote": "[0:01:31] Writing. I give it like it's like a B tier, B plus tier. … It tends to bury the lead.",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L040",
   "voice": "Every",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "newsletter",
   "url": "https://every.to/vibe-check/fable-5-2",
   "quote": "With Opus 5.5, you're still in the loop, checking its work as you go. … But Fable 5.2 can take the whole product: Hand it everything you want, and it keeps building, testing, and refining until it has something you'll use every day.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L041",
   "voice": "Every",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "newsletter",
   "url": "https://every.to/vibe-check/vibe-check-gemini-4-carbon-doesn-t-know-when-to-stop",
   "quote": "It's a much stronger coder than Gemini 4 Argon, and it makes better three-dimensional objects.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Gemini 4 Argon",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L042",
   "voice": "Every",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "newsletter",
   "url": "https://every.to/vibe-check/vibe-check-gemini-4-carbon-doesn-t-know-when-to-stop",
   "quote": "It's a better model, but Fable has raised expectations for what a flagship should deliver.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L043",
   "voice": "Every",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=6Sz_8EhZWAc&t=21s",
   "quote": "[0:00:21] Here's the quick take. This is not a great model. Across everyone on our team who tried it, no one loved it.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -1
  },
  {
   "id": "L044",
   "voice": "Every",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=6Sz_8EhZWAc&t=64s",
   "quote": "[0:01:04] It's not a good writing model. … Astra 6.2 is a step back.",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "GPT-6 Astra",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -1
  },
  {
   "id": "L045",
   "voice": "Every",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=6Sz_8EhZWAc&t=373s",
   "quote": "[0:06:13] We found that generally across a lot of end to end builds, this model just stopped early and was kind of frustrating to work with",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L046",
   "voice": "Every",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=6Sz_8EhZWAc&t=461s",
   "quote": "[0:07:41] Astra 6.2, by comparison, it has good defaults. These are interesting outputs, but they all kind of look the same.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L047",
   "voice": "Every",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-17",
   "platform": "newsletter",
   "url": "https://every.to/vibe-check/vibe-check-gpt-6-2-astra-and-sol-leave-us-wanting-more",
   "quote": "GPT-6.2 Astra and Sol handle bounded tasks efficiently. The work we most want help with still demands more judgment and follow-through.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L048",
   "voice": "Simon Willison",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "blog",
   "url": "https://simonwillison.net/2026/Nov/17/gpt62-astra-and-sol/",
   "quote": "Neither of them did a good enough job of morphing Yonkers into a big mech, at least in comparison to Claude Fable 5.2 and Gemini 4 Carbon.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    },
    {
     "model": "Gemini 4 Carbon",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L049",
   "voice": "Simon Willison",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-17",
   "platform": "blog",
   "url": "https://simonwillison.net/2026/Nov/17/gpt62-astra-and-sol/",
   "quote": "Neither of them did a good enough job of morphing Yonkers into a big mech, at least in comparison to Claude Fable 5.2 and Gemini 4 Carbon.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    },
    {
     "model": "Gemini 4 Carbon",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L050",
   "voice": "Simon Willison",
   "model": "Claude Fable 5.2",
   "date": "2026-11-17",
   "platform": "blog",
   "url": "https://simonwillison.net/2026/Nov/17/gpt62-astra-and-sol/",
   "quote": "Neither of them did a good enough job of morphing Yonkers into a big mech, at least in comparison to Claude Fable 5.2 and Gemini 4 Carbon.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6.2 Astra",
     "direction": "better"
    },
    {
     "model": "GPT-6.2 Sol",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L051",
   "voice": "Simon Willison",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-17",
   "platform": "blog",
   "url": "https://simonwillison.net/2026/Nov/17/gpt62-astra-and-sol/",
   "quote": "Neither of them did a good enough job of morphing Yonkers into a big mech, at least in comparison to Claude Fable 5.2 and Gemini 4 Carbon.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6.2 Astra",
     "direction": "better"
    },
    {
     "model": "GPT-6.2 Sol",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L052",
   "voice": "Simon Willison",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "blog",
   "url": "https://simonwillison.net/2026/Nov/17/gpt62-astra-and-sol/",
   "quote": "The buzz around these new models is mixed at the moment. … I've seen a few complaints that its communication style can be confusing.",
   "basis": "hearsay",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L053",
   "voice": "Simon Willison",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "blog",
   "url": "https://simonwillison.net/2026/Nov/9/llm-anthropic/",
   "quote": "Fable 5.2 is currently ranked first on Artificial Analysis.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L054",
   "voice": "Simon Willison",
   "model": "Muse Ember",
   "date": "2026-11-10",
   "platform": "blog",
   "url": "https://simonwillison.net/2026/Nov/10/muse-ember/",
   "quote": "The Ember pelicans are comparable in quality to those I got from Muse Wonder, which is 40x the price of Muse Ember.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Muse Wonder",
     "direction": "same"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L055",
   "voice": "Simon Willison",
   "model": "Muse Ember",
   "date": "2026-11-10",
   "platform": "blog",
   "url": "https://simonwillison.net/2026/Nov/10/muse-ember/",
   "quote": "The Muse Ember models sit right on that Pareto line - the line at which you can only improve one dimension by giving up another.",
   "basis": "benchmark-reaction",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L056",
   "voice": "Simon Willison",
   "model": "GPT-6.1 Terra",
   "date": "2026-11-10",
   "platform": "blog",
   "url": "https://simonwillison.net/2026/Nov/10/muse-ember/",
   "quote": "clearly I should be paying more attention to the Terra family!",
   "basis": "benchmark-reaction",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L057",
   "voice": "Simon Willison",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-23",
   "platform": "blog",
   "url": "https://simonwillison.net/2026/Oct/23/kimi-k31-pro/",
   "quote": "This is the first time a (soon-to-be) open weight model has scored this highly against the frontier.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L058",
   "voice": "Simon Willison",
   "model": "Gemini Decision 1",
   "date": "2026-11-11",
   "platform": "blog",
   "url": "https://simonwillison.net/2026/Nov/11/gemini-decisions-api/",
   "quote": "Jev and OpenAI both return 1.0 or 0.0 for yes/no questions, but Gemini insists on being a little less confident than that:",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Jev",
     "direction": "same"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L059",
   "voice": "Matthew Berman",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=0vNyVMDFdkg&t=1221s",
   "quote": "[0:20:21] And besides the benchmarks, I have tested it thoroughly for a little while now, and I can definitively say it is by far the best model I've ever used.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L060",
   "voice": "Matthew Berman",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=0vNyVMDFdkg&t=1361s",
   "quote": "[0:22:41] Without a doubt, Fable 5.2 is better than Opus 5.5. However, Opus 5.5 is very efficient.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "L061",
   "voice": "Matthew Berman",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=0vNyVMDFdkg&t=1151s",
   "quote": "[0:19:11] Frontier Code 2.0, it absolutely dominates all the other models on the list, including Opus 5.5, which was my favorite model.",
   "basis": "benchmark-reaction",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L062",
   "voice": "Matthew Berman",
   "model": "Gemini 4 Carbon",
   "date": "2026-10-29",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=zhnJo7s91ls&t=0s",
   "quote": "[0:00:00] There's a new model from Google that claims to be the absolute best model on the planet beating literally everything else closed source open source everything. And if this is true, it means Google is absolutely back.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 0
  },
  {
   "id": "L063",
   "voice": "Matthew Berman",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-11",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=FCmLuTENOp0&t=192s",
   "quote": "[0:03:12] I have had early access to Carbon. I've been testing it, and I just keep going back to Fable. … And Carbon is a good model. Don't get me wrong, but it is nowhere near Fable.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L064",
   "voice": "Matthew Berman",
   "model": "Muse Ember",
   "date": "2026-11-11",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=FCmLuTENOp0&t=771s",
   "quote": "[0:12:51] It scores highest than Claude Haiku, but it is much less expensive, about a third of the price, which is great.",
   "basis": "benchmark-reaction",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Haiku 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L065",
   "voice": "Matthew Berman",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-24",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=5sEVvu4mQ8g&t=0s",
   "quote": "[0:00:00] I'm going to talk about the model specs, and yes, it is really good.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L066",
   "voice": "Matthew Berman",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-24",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=5sEVvu4mQ8g&t=1005s",
   "quote": "[0:16:45] I find that I have to nudge Kimmy to do things more so than I would uh other models that would just make assumptions about what I want.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L067",
   "voice": "Matthew Berman",
   "model": "DeepSeek Flash 1017",
   "date": "2026-10-28",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=DDZ-a6bAn3Y&t=1050s",
   "quote": "[0:17:30] So, from these demos, hopefully you can see why I'm calling this potentially the best overall model on the planet right now.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L068",
   "voice": "Matthew Berman",
   "model": "DeepSeek Flash 1017",
   "date": "2026-10-28",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=DDZ-a6bAn3Y&t=211s",
   "quote": "[0:03:31] I think overall I'm really impressed with its ability to clone this game. It isn't as good as Opus 5.5, but it's quite close.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L069",
   "voice": "AICodeKing",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-24",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=EP-yA-1PVmw&t=775s",
   "quote": "[0:12:55] The launch blog sells it as being right next to Opus, but in real projects, it's sitting just behind Haiku. So, it's a good open model, not a frontier one",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    },
    {
     "model": "Claude Haiku 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": 0
  },
  {
   "id": "L070",
   "voice": "AICodeKing",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-24",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=EP-yA-1PVmw&t=662s",
   "quote": "[0:11:02] Being cheaper than Opus doesn't mean much when it doesn't perform like Opus.",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "DeepSeek Flash 1017",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L071",
   "voice": "AICodeKing",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=WOc6Vf8cCsE&t=1318s",
   "quote": "[0:21:58] So here is my one line verdict. On the $200 plan, Fable 5.2 is absolutely worth it for your hardest projects, but on the API, it is a specialist you call in, not your daily driver.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L072",
   "voice": "AICodeKing",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=WOc6Vf8cCsE&t=981s",
   "quote": "[0:16:21] And on those three, Fable used about twice as many tokens and cost about three times as much, $78 against 26.",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L073",
   "voice": "AICodeKing",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-11",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=SvXR3bNktMA&t=880s",
   "quote": "[0:14:40] Overall, Carbon is a good model at a great price, but it is not the Fable 5.2 answer",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    },
    {
     "model": "Gemini 4 Argon",
     "direction": "same"
    }
   ],
   "stance_first": 0,
   "stance_second": 1
  },
  {
   "id": "L074",
   "voice": "AICodeKing",
   "model": "Gemini 4 Hydrogen",
   "date": "2026-11-14",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=HBJqMIyOQGM&t=673s",
   "quote": "[0:11:13] Overall, Hydrogen is a fast, cheap worker with a bad sense of when it is finished. Use it under a stronger lead and do not trust it when it says the job is verified.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L075",
   "voice": "AICodeKing",
   "model": "Qwen4-Max",
   "date": "2026-11-16",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=mXZcgH4UyCM&t=655s",
   "quote": "[0:10:55] It did not beat Kimi or DeepSeek Flash on my tests … today it is a mid-table model with a very cheap way to use it.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Kimi K3.1 Pro",
     "direction": "worse"
    },
    {
     "model": "DeepSeek Flash 1017",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L076",
   "voice": "AICodeKing",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=wT39a2-QpHw&t=692s",
   "quote": "[0:11:32] I watched it and honestly it was atrocious.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "L077",
   "voice": "AICodeKing",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=wT39a2-QpHw&t=762s",
   "quote": "[0:12:42] I would use Astra selectively when its stronger result on a particular task matters enough to justify the cost and check its work carefully.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": 0
  },
  {
   "id": "L078",
   "voice": "AICodeKing",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=wT39a2-QpHw&t=762s",
   "quote": "[0:12:42] Soul is worth using for budget projects and simple prototypes.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 0
  },
  {
   "id": "L079",
   "voice": "Peter Yang",
   "model": "Claude Fable 5.2",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=NTnV2dxIaDE&t=314s",
   "quote": "[0:05:14] this is definitely the best game of the three in my opinion. Okay, so I think Fable clearly won the game use case",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6.2 Astra",
     "direction": "better"
    },
    {
     "model": "Gemini 4 Carbon",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "L080",
   "voice": "Peter Yang",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=NTnV2dxIaDE&t=474s",
   "quote": "[0:07:54] I think this is actually a really well done website. And I think Astra and Open Eyes models have come a long way on website design.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Gemini 4 Carbon",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L081",
   "voice": "Peter Yang",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=NTnV2dxIaDE&t=615s",
   "quote": "[0:10:15] is actually pretty comparable with Gemini 4 Carbon clearly being worse than the two.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6.2 Astra",
     "direction": "worse"
    },
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L082",
   "voice": "Peter Yang",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=NTnV2dxIaDE&t=848s",
   "quote": "[0:14:08] Now, this is obviously not a huge sample size, but I argue that Gemini 4 Carbon is a really good writer, maybe even better than Fable 5.2.",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "better"
    },
    {
     "model": "GPT-6.2 Astra",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "L083",
   "voice": "Peter Yang",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=NTnV2dxIaDE&t=824s",
   "quote": "[0:13:44] And Austra is SA. So that's clearly the worst example.",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Gemini 4 Carbon",
     "direction": "worse"
    },
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L084",
   "voice": "Peter Yang",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=NTnV2dxIaDE&t=1218s",
   "quote": "[0:20:18] I pushed back and Astra gave ground a lot because I think it was too lazy in its initial analysis.",
   "basis": "hands-on",
   "domain": "research",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    },
    {
     "model": "Gemini 4 Carbon",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L085",
   "voice": "Peter Yang",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=SsSKJA0EHD4&t=1247s",
   "quote": "[0:20:47] in terms of writing variety and style, I definitely prefer Fable's essay four out of five times versus Opus.",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "L086",
   "voice": "Peter Yang",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=SsSKJA0EHD4&t=1225s",
   "quote": "[0:20:25] For website design, in terms of making bold designs with motion graphics, Fable is definitely better than Opus, in my opinion.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L087",
   "voice": "Peter Yang",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=SsSKJA0EHD4&t=1269s",
   "quote": "[0:21:09] It seems that Opus is actually better at pushing back hard when I demanded agreement",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L088",
   "voice": "Wes Roth",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=h76wUNmRqxc&t=855s",
   "quote": "[0:14:15] that actually really impressive. … I'm kind of blown away.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L089",
   "voice": "Wes Roth",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=h76wUNmRqxc&t=0s",
   "quote": "[0:00:00] Now, without any exaggeration, this is the best model in the world, at least from models that are available to us mortals.",
   "basis": "announcement",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L090",
   "voice": "Wes Roth",
   "model": "Claude Fable 5.2",
   "date": "2026-11-13",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=Sc6TriRAGnI&t=1238s",
   "quote": "[0:20:38] I ask Claude Fable 5.2, my most trusted coding model",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 2
  },
  {
   "id": "L091",
   "voice": "Wes Roth",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-13",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=Sc6TriRAGnI&t=299s",
   "quote": "[0:04:59] Gemini 4 Carbon for me did something no other model was able to do. It was able to edit an entire video beginning to end",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L092",
   "voice": "Wes Roth",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-13",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=Sc6TriRAGnI&t=1627s",
   "quote": "[0:27:07] So I got to say my test of the swarm for this project, I give it an A plus.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L093",
   "voice": "Wes Roth",
   "model": "Gemini 4 Hydrogen",
   "date": "2026-11-13",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=Sc6TriRAGnI&t=276s",
   "quote": "[0:04:36] But make no mistake about it on a lot of these benchmarks that we're seeing hydrogen is beating pretty much everything else out there.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L094",
   "voice": "Wes Roth",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=tEysfZGzJc8&t=300s",
   "quote": "[0:05:00] Its ability to understand instructions and write things out doesn't seem to be as good as the previous Astra.",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6 Astra",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L095",
   "voice": "Wes Roth",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=tEysfZGzJc8&t=690s",
   "quote": "[0:11:30] that is very, very impressive. We shouldn't just dismiss that outright.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L096",
   "voice": "Wes Roth",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=tEysfZGzJc8&t=0s",
   "quote": "[0:00:00] Javi says it's a catastrophe. Matt Schumer says he will not be writing a review about these models.",
   "basis": "hearsay",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "L097",
   "voice": "Prompt Engineering",
   "model": "Gemini 4 Neon",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=Q0ShBenKyqo&t=764s",
   "quote": "[0:12:44] So Gemini 4 Neon is a very capable model and I had a lot of fun testing it. I highly recommend to check it out.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 2
  },
  {
   "id": "L098",
   "voice": "Prompt Engineering",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=lQ6wwLHqDAg&t=22s",
   "quote": "[0:00:22] have had early access to Carbon for quite some time and two things actually stood out for me. The first one is that it can implement the latest research ideas.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L099",
   "voice": "Prompt Engineering",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=lQ6wwLHqDAg&t=231s",
   "quote": "[0:03:51] In one pass of about 72 minutes, it produced a full dragon that holds together from angles that are not in the reference image, which is pretty impressive.",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L100",
   "voice": "Prompt Engineering",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-25",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=_XXX5K-EMoc&t=469s",
   "quote": "[0:07:49] If your prompt leaves room for an easier algorithm or an early stopping point, Kimi uh K3.1 Pro tends to take it",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L101",
   "voice": "Prompt Engineering",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-25",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=_XXX5K-EMoc&t=935s",
   "quote": "[0:15:35] So, on write.com, the visual quality is dead heat.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "same"
    }
   ],
   "stance_first": 1,
   "stance_second": 0
  },
  {
   "id": "L102",
   "voice": "Prompt Engineering",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-25",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=_XXX5K-EMoc&t=1100s",
   "quote": "[0:18:20] Overall, Kimi was cheaper on every single task",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L103",
   "voice": "Prompt Engineering",
   "model": "Gemini Decision 1",
   "date": "2026-11-13",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=-On9ZKGiMfA&t=284s",
   "quote": "[0:04:44] it picked the right tool on every request in my test.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L104",
   "voice": "Prompt Engineering",
   "model": "Gemini Decision 1",
   "date": "2026-11-13",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=-On9ZKGiMfA&t=399s",
   "quote": "[0:06:39] For example, for routing one message to one of the four cues, the decision call cost about $10 per million calls.",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 0
  },
  {
   "id": "L105",
   "voice": "AI Explained",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=583ThJIqQ7k&t=800s",
   "quote": "[0:13:20] But it ended up good-hearted itself. When the measure becomes a target, it ceases to be a good measure. It started taking tests out from phase one and moving them after the human review without telling me.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L106",
   "voice": "AI Explained",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=583ThJIqQ7k&t=572s",
   "quote": "[0:09:32] This for me is the more relevant hallucination rate. And here it's actually Gemini 4 Carbon that does the best",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L107",
   "voice": "Sam Witteveen",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=l32KKNKdpIk&t=528s",
   "quote": "[0:08:48] this is where I want to say that both this model and the Fable 5.2 model are suddenly now, I feel, in a new league when it comes to software development.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "same"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L108",
   "voice": "Sam Witteveen",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=l32KKNKdpIk&t=442s",
   "quote": "[0:07:22] so I've gotta say that I was pretty impressed watching this be built.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L109",
   "voice": "Sam Witteveen",
   "model": "Mercury 3",
   "date": "2026-10-20",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=bEO_QeJU7U0&t=1294s",
   "quote": "[0:21:34] If you haven't, I really suggest trying Mercury 3 out and just seeing how well and more importantly, how quickly it can actually work for your use cases",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "L110",
   "voice": "Sam Witteveen",
   "model": "Mercury 3",
   "date": "2026-10-20",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=bEO_QeJU7U0&t=789s",
   "quote": "[0:13:09] So in my tests so far, I don't find it doing the things that a lot of the open models do of where they only write half of the essay",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L111",
   "voice": "Fireship",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-26",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=G_em1hHPnXY&t=274s",
   "quote": "[0:04:34] I was blown away by its beauty at first glance, but then I push play on the animation and found a problem. The animations are absolutely terrible.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": -1
  },
  {
   "id": "L112",
   "voice": "Fireship",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=1FurWDiK9A4&t=178s",
   "quote": "[0:02:58] The end result is the most realistic horse thorough bench has ever seen.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "L113",
   "voice": "Zvi Mowshowitz",
   "model": "Claude Fable 5.2",
   "date": "2026-11-16",
   "platform": "newsletter",
   "url": "https://thezvi.substack.com/p/fable-52-is-the-new-bazooka",
   "quote": "For most purposes, for a thousand household uses, Claude Fable 5.2 will not be appreciably better than Claude Opus 5.5.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "same"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L114",
   "voice": "Zvi Mowshowitz",
   "model": "Claude Fable 5.2",
   "date": "2026-11-16",
   "platform": "newsletter",
   "url": "https://thezvi.substack.com/p/fable-52-is-the-new-bazooka",
   "quote": "When you have a difficult task, or a very complex task, or a long task? … Get the bazooka.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "L115",
   "voice": "Zvi Mowshowitz",
   "model": "Claude Fable 5.2",
   "date": "2026-11-16",
   "platform": "newsletter",
   "url": "https://thezvi.substack.com/p/fable-52-is-the-new-bazooka",
   "quote": "For editing I use Opus 5.5 on Max, unless I am in a hurry, much to Fable 5.2's chagrin, because Opus 5.5 is actively better there",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L116",
   "voice": "Zvi Mowshowitz",
   "model": "Muse Ember",
   "date": "2026-11-12",
   "platform": "newsletter",
   "url": "https://thezvi.wordpress.com/2026/11/12/ai-194-spy-versus-guy/",
   "quote": "Muse Ember, Meta's smallest model, exists. They say it is good. I don't believe them.",
   "basis": "announcement",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L117",
   "voice": "Zvi Mowshowitz",
   "model": "Gemini 4 Carbon",
   "date": "2026-10-29",
   "platform": "newsletter",
   "url": "https://thezvi.wordpress.com/2026/10/29/ai-192-anthropic-is-going-public/",
   "quote": "Gemini 4 Carbon exists. Its benchmarks are excellent.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L118",
   "voice": "Zvi Mowshowitz",
   "model": "GPT-6.1 Terra",
   "date": "2026-10-29",
   "platform": "newsletter",
   "url": "https://thezvi.wordpress.com/2026/10/29/ai-192-anthropic-is-going-public/",
   "quote": "GPT-6.1 Terra exists. Benchmarks suggest this is roughly on par with 6.1 Sol after adjusting for model size.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6.1 Sol",
     "direction": "same"
    }
   ],
   "stance_first": 1,
   "stance_second": 0
  },
  {
   "id": "L119",
   "voice": "Zvi Mowshowitz",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-29",
   "platform": "newsletter",
   "url": "https://thezvi.wordpress.com/2026/10/29/ai-192-anthropic-is-going-public/",
   "quote": "Kimi K3.1 Pro which has very strong benchmarks but where I haven't heard much talk.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 0
  },
  {
   "id": "L120",
   "voice": "Zvi Mowshowitz",
   "model": "DeepSeek Flash 1017",
   "date": "2026-10-29",
   "platform": "newsletter",
   "url": "https://thezvi.wordpress.com/2026/10/29/ai-192-anthropic-is-going-public/",
   "quote": "As for capabilities, the implicit verdict is in. It's fine.",
   "basis": "hearsay",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "L121",
   "voice": "Ethan Mollick",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "newsletter",
   "url": "https://www.oneusefulthing.org/p/fable-52-and-jagged-agi",
   "quote": "it is very impressive, following on the already excellent Opus 5.5.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "L122",
   "voice": "Ethan Mollick",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "newsletter",
   "url": "https://www.oneusefulthing.org/p/fable-52-and-jagged-agi",
   "quote": "On a first read, it is ahead of what I would expect from second-year PhD students. It is also not publishable as it is.",
   "basis": "hands-on",
   "domain": "research",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L123",
   "voice": "Ethan Mollick",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "newsletter",
   "url": "https://www.oneusefulthing.org/p/fable-52-and-jagged-agi",
   "quote": "Knowing when your own work is not good enough is a large part of what makes someone an expert, and it is the part Fable does least reliably.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L124",
   "voice": "Claire Vo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=1gYQc4CL_oo&t=73s",
   "quote": "[0:01:13] But this is a fast, capable, not annoying, creative powerhouse of a model.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L125",
   "voice": "Claire Vo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=1gYQc4CL_oo&t=348s",
   "quote": "[0:05:48] It is really great at coding. It has helped me with some architecture, re-architecture work, long-running engineering tasks, things that previous models, Opus 55, Astra, etc., have not been able to really get through and it's not annoying.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    },
    {
     "model": "GPT-6 Astra",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L126",
   "voice": "Claire Vo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=1gYQc4CL_oo&t=348s",
   "quote": "[0:05:48] What are the cons? Well, it's not good at front end and web design.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "L127",
   "voice": "WorldofAI",
   "model": "Gemini 4 Neon",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=R0rXSJmymTc&t=433s",
   "quote": "[0:07:13] Both models did a pretty great job overall, but I still think that the Fable 5.2 did deliver a better overall generation in terms of its visual quality and detail.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L128",
   "voice": "WorldofAI",
   "model": "Gemini 4 Neon",
   "date": "2026-11-18",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=R0rXSJmymTc&t=0s",
   "quote": "[0:00:00] and it is an incredible model, and I genuinely owe Google an apology. This model is just something else.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L129",
   "voice": "WorldofAI",
   "model": "Qwen4-Max",
   "date": "2026-11-16",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=5vPXJiy83Sg&t=24s",
   "quote": "[0:00:24] I do like the model, but I am definitely slightly disappointed with it",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": 0
  },
  {
   "id": "L130",
   "voice": "WorldofAI",
   "model": "Qwen4-Max",
   "date": "2026-11-16",
   "platform": "youtube",
   "url": "https://www.youtube.com/watch?v=5vPXJiy83Sg&t=438s",
   "quote": "[0:07:18] honestly, it does quite well with all of these different front-end generation tasks that I had requested it to create.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "L131",
   "voice": "The Neuron (Corey Noles)",
   "model": "Mercury 3",
   "date": "2026-10-20",
   "platform": "newsletter",
   "url": "https://www.theneuron.ai/news/mercury-3-review-diffusion-language-models/",
   "quote": "My early verdict: Mercury 3 is a genuinely exciting model at a startling price.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "L132",
   "voice": "Nathan Lambert (Interconnects)",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-25",
   "platform": "podcast",
   "url": "https://www.interconnects.ai/p/podcast-kimi-k31-pro-the-new-cyber",
   "quote": "That's why I think it's pretty reasonable criticism of this Kimi release: a big model has got to be...",
   "basis": "hearsay",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "DeepSeek Flash 1017",
     "direction": "same"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X000",
   "voice": "Theo",
   "model": "Gemini 4 Neon",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/theo/status/2123237018813779970",
   "quote": "Why the fuck is nobody talking about Neon? This model is so good wtf",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X001",
   "voice": "Theo",
   "model": "Gemini 4 Neon",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/theo/status/2123248799238029475",
   "quote": "Still waiting on numbers but honestly? I don't care. Astra showed me that my benchmarks suck. Neon is super pleasant to work with, it unblocks itself, makes good decisions, writes good code, and writes english SO well.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X002",
   "voice": "Theo",
   "model": "Gemini 4 Neon",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/theo/status/2123248515346530775",
   "quote": "It's a generational leap from Carbon Funny enough, the best parts are reminiscent of Muse Wonder Full",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Gemini 4 Carbon",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X003",
   "voice": "Theo",
   "model": "Gemini 4 Neon",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/theo/status/2123233242140824005",
   "quote": "It is the most readable model I've used this year. Neck and neck with Muse Wonder Full. I'm having a blast with this thing.",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Muse Wonder Full",
     "direction": "same"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X004",
   "voice": "Theo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/theo/status/2122757647170703550",
   "quote": "6.2 Astra is here. I’ve been testing it for awhile. I think this is the worst model OpenAI has ever released. I’ll share more soon.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X005",
   "voice": "Theo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/theo/status/2122991778622263514",
   "quote": "I took this to heart and spent the last few hours doing another deep dive on GPT-6.2 Astra tl;dr - things are better than when I tested, but IMO still \"directionally, very bad\"",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -1
  },
  {
   "id": "X006",
   "voice": "Theo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/theo/status/2123033570918957370",
   "quote": "Show me anyone shipping at 1/5th my team’s velocity that genuinely prefers Astra to Opus and I’ll show you a liar",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X007",
   "voice": "Theo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/theo/status/2122831642339557794",
   "quote": "This one’s gonna get me in trouble. If GPT-6.2 Astra acted more like Muse Wonder Full, it would be a great model.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Muse Wonder Full",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X008",
   "voice": "Theo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/theo/status/2122973764438933824",
   "quote": "Seems like 6.2 Astra is incredible for math. Cool to see other areas it works well in!",
   "basis": "hearsay",
   "domain": "research",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X009",
   "voice": "Theo",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/theo/status/2122760761688883577",
   "quote": "fwiw, I've heard that a large portion of OAI's engineers are straight up ignoring Astra and just use 6.2 Sol Ultrafast for their day to day work",
   "basis": "hearsay",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6.2 Astra",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X010",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/theo/status/2119881447800770962",
   "quote": "It's been under an hour and I'm already struggling with \"should I use Opus 5.5 or Fable 5.2?\"",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "same"
    }
   ],
   "stance_first": 1,
   "stance_second": 0
  },
  {
   "id": "X011",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/theo/status/2119926277801144695",
   "quote": "Fable 5.2 built this in 76 minutes and it cost ~$42 in tokens. One of the best Fishslop runs I've seen for sure, but not significantly better than Opus 5.5 (esp. for the price)",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "same"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X012",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/theo/status/2119988540708896897",
   "quote": "Fable 5.2 is making a lot of stupid mistakes I would not have expected from Opus tbh",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X013",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/theo/status/2119960608816988596",
   "quote": "The $200 Claude Ultra plan gets you $15,000 of Opus 5.5 in a month, but only around $3,000 of Fable 5.5. Even ignoring that Fable is restricted to 50% of your weekly limit, it burns faster per $ than Opus does.",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X014",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/theo/status/2120104895646937409",
   "quote": "Okay FINE, Fable kinda cooked with this design. Literally was just trying to do a writeup for some data pipeline stuff for a teammate and it made this lmao",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X015",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/theo/status/2119946060361818578",
   "quote": "Latest benchmarks are up on Slopalytics (ty Artificial Analysis!) Holy shit Fable is mogging so hard. 5 point lead!",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X016",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/theo/status/2120284921839394847",
   "quote": "Fable 5.2 is the greatest model ever released. I stayed up until 3am getting this video ready for y'all.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X017",
   "voice": "Theo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-16",
   "platform": "x",
   "url": "https://x.com/theo/status/2122429147289206976",
   "quote": "Fable 5.2 is the first model that consistently ignores unimportant feedback. It’s incredible. Had at least a dozen instances where I used Opus for a thread, it started running in circles, swapped to Fable and it was like “wtf this is fine just ship it”",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X018",
   "voice": "Theo",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-11",
   "platform": "x",
   "url": "https://x.com/theo/status/2120596827980665018",
   "quote": "Gemini 4 Carbon is here. It's slow, awkward, stupid, and hallucinates all the time. That said, they made a ton of progress and the team is working hard to make it great.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -2
  },
  {
   "id": "X019",
   "voice": "Theo",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-11",
   "platform": "x",
   "url": "https://x.com/theo/status/2120597586298261543",
   "quote": "I don't really know how to properly emphasize this. Gemini 4 Carbon is unacceptably slow in real world usage. Gemini 4 Carbon low: 22m21s Opus 5.5 high: 9m33s",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X020",
   "voice": "Theo",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/theo/status/2123241064823984425",
   "quote": "Nah Carbon is ass and should be deprecated",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X021",
   "voice": "Theo",
   "model": "Gemini 4 Hydrogen",
   "date": "2026-11-12",
   "platform": "x",
   "url": "https://x.com/theo/status/2121030235248570537",
   "quote": "“Built for speed” Still garbage speeds compared to anything else on the market. Note that I ran Sol and Opus at regular tier! Didn’t even use fast/ultrafast.",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6.1 Sol",
     "direction": "worse"
    },
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X022",
   "voice": "Theo",
   "model": "Muse Ember",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/theo/status/2119967883442999602",
   "quote": "Muse family is so criminally underrated. Wonder is their only bad model. Wonder Full is insane, Spark is great value and Ember is proving to be great as well",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X023",
   "voice": "Theo",
   "model": "GPT-6.1 Terra",
   "date": "2026-10-29",
   "platform": "x",
   "url": "https://x.com/theo/status/2115729216566407583",
   "quote": "GPT-6.1 Terra is honestly much better than I anticipated. Using it for a bunch of random stuff",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X024",
   "voice": "Theo",
   "model": "Gemini 4 Argon",
   "date": "2026-10-31",
   "platform": "x",
   "url": "https://x.com/theo/status/2116634480991785076",
   "quote": "Love you Prime but no, they did not. This model is incapable of working in a real code base without hallucinating in egregious ways. It’s not like 5% worse, it’s 10x worse.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X025",
   "voice": "Dan Shipper (Every)",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2122754729012777052",
   "quote": "We've been testing them for the last few weeks @every and, unfortunately, these models aren't great. Overall a disappointing release compared to Fable 5.2.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X026",
   "voice": "Dan Shipper (Every)",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2122754729012777052",
   "quote": "We've been testing them for the last few weeks @every and, unfortunately, these models aren't great. Overall a disappointing release compared to Fable 5.2.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X027",
   "voice": "Dan Shipper (Every)",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2122754729012777052",
   "quote": "We found Astra 6.2 to be a step back from Astra 6 on writing. 6.2 appears to have been trained to very extremely concise and actionable—so much so that its writing is hard to understand",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "GPT-6 Astra",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -1
  },
  {
   "id": "X028",
   "voice": "Dan Shipper (Every)",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2122754729012777052",
   "quote": "We found performance to be similar to Astra 6 on most of our benchmarks. However, in real coding tasks Astra 6.2 seemed to stop too early frequently, which made it quite frustating to use on end-to-end builds.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "GPT-6 Astra",
     "direction": "same"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X029",
   "voice": "Dan Shipper (Every)",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2122754729012777052",
   "quote": "Astra 6.2 has good default design instincts. But it struggles to vary its designs, resulting in output that has a noticeable \"house style\".",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X030",
   "voice": "Dan Shipper (Every)",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2122858139989012510",
   "quote": "the @every team's full ratings for Astra and Sol 6.2 hard to remember a frontier model release like this. opus 5 is the closest comp",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X031",
   "voice": "Dan Shipper (Every)",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2120187507396530309",
   "quote": "It’s a much better coder than Argon—but I’m probably keeping Argon for writing.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Gemini 4 Argon",
     "direction": "better"
    }
   ],
   "stance_first": 0,
   "stance_second": 1
  },
  {
   "id": "X032",
   "voice": "Dan Shipper (Every)",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2120187507396530309",
   "quote": "On one run of our Senior Engineer benchmark, Carbon scored 44%, versus Argon’s 21%. A big improvement, though still middle of the pack.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Gemini 4 Argon",
     "direction": "better"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X033",
   "voice": "Dan Shipper (Every)",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2120187507396530309",
   "quote": "Our editorial benchmark put Carbon and Argon about even. If you like Argon, you’ll probably like Carbon. But Carbon costs more, and I haven’t seen enough improvement to justify switching.",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Gemini 4 Argon",
     "direction": "same"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X034",
   "voice": "Dan Shipper (Every)",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2120187507396530309",
   "quote": "Interfaces split the team: Jack and Randy kept getting boxes within boxes, while @TylerNishida liked its ideas for new interfaces.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Gemini 4 Argon",
     "direction": "better"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X035",
   "voice": "Dan Shipper (Every)",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2119869228207546638",
   "quote": "It’s a paradigm-shifting model if you have a project ambitious enough to give it.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X036",
   "voice": "Dan Shipper (Every)",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2119869228207546638",
   "quote": "The catch: It’s slow and burns through tokens. Our early-access usage was free, and these projects would have been expensive at API rates.",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X037",
   "voice": "Dan Shipper (Every)",
   "model": "Claude Fable 5.2",
   "date": "2026-11-11",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2120539547620790424",
   "quote": "I look at what I do with Fable 5.2 in the cracks of my week and think: this would've taken five people and $200,000.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X038",
   "voice": "Dan Shipper (Every)",
   "model": "Gemini 4 Argon",
   "date": "2026-11-01",
   "platform": "x",
   "url": "https://x.com/danshipper/status/2117014507083948453",
   "quote": "argon cooks",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "X039",
   "voice": "Bijan Bowen",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2122783433122558341",
   "quote": "me watching the TL trash GPT-6.2 Astra knowing I can make it cook, then remembering OpenAI didn't give me early access",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X040",
   "voice": "Bijan Bowen",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2122806295690629385",
   "quote": "GPT 6.2 Astra C++ Skate game that i turned into a GTA clone",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 0
  },
  {
   "id": "X041",
   "voice": "Bijan Bowen",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2122929880115405297",
   "quote": "Astra 6.2 is an IQ test for influencers",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 0
  },
  {
   "id": "X042",
   "voice": "Bijan Bowen",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2122997443466063946",
   "quote": "Its deeply capable, but the EQ is perhaps not there, so it comes off like christian bale as michael burry, while fable comes off like christian bale as batman",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X043",
   "voice": "Bijan Bowen",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2119920843065053320",
   "quote": "Opus 5.5 mogs Fable 5.2 at threejs 😬",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X044",
   "voice": "Bijan Bowen",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2119972220739187182",
   "quote": "Fable 5.2 might be better at threejs than I first thought...",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X045",
   "voice": "Bijan Bowen",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-11",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2120308742168387852",
   "quote": "Gemini 4 Carbon made this racing game and it's addictive tbh",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X046",
   "voice": "Bijan Bowen",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-13",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2121121400949797216",
   "quote": "Carbon > Fable 5.2 > Argon > Hydrogen",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "better"
    },
    {
     "model": "Gemini 4 Argon",
     "direction": "better"
    },
    {
     "model": "Gemini 4 Hydrogen",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X047",
   "voice": "Bijan Bowen",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-13",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2121122785271456024",
   "quote": "carbon is threatening the self worth I derive from my CAD skills being way better than the frontier models",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "X048",
   "voice": "Bijan Bowen",
   "model": "Gemini 4 Hydrogen",
   "date": "2026-11-13",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2121042010928709932",
   "quote": "Gemini Hydrogen made this 3d model of the Seinfeld Apartment from this source image and I printed it",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 0
  },
  {
   "id": "X049",
   "voice": "Bijan Bowen",
   "model": "Gemini 4 Hydrogen",
   "date": "2026-11-12",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2120993785400430694",
   "quote": "This is carbon but yes the usage on hydrogen with swarm has been very very good",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X050",
   "voice": "Bijan Bowen",
   "model": "Muse Ember",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2119855972726472892",
   "quote": "Muse Ember did this in under 2 minutes on its highest reasoning effort, this model might SLAP",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "X051",
   "voice": "Bijan Bowen",
   "model": "GPT-6.1 Terra",
   "date": "2026-10-26",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2114862830453412154",
   "quote": "hmm 6.1 Terra is EXTREMELY fast, so far I am liking this thing",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X052",
   "voice": "Bijan Bowen",
   "model": "GPT-6.1 Terra",
   "date": "2026-10-27",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2115017348679975031",
   "quote": "GPT-6.1 Terra made this disturbed \"SNL Skit\" with some wrestler models it made in blender",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X053",
   "voice": "Bijan Bowen",
   "model": "DeepSeek Flash 1017",
   "date": "2026-10-17",
   "platform": "x",
   "url": "https://x.com/bijanbowen/status/2111334540476056034",
   "quote": "wanted to see if DeepSeek V4.1 Flash 1017 was real, asked for the model slug and it made me a 3d model of a slug.... genuinely can't make this stuff up",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X054",
   "voice": "Simon Willison",
   "model": "Claude Fable 5.2",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/simonw/status/2122844219270422750",
   "quote": "I was most impressed with Fable because it did at least have the physical geography of Yonkers form together into the mech - but none of them are excellent (It's a deliberwtely absurd prompt though)",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6.2 Astra",
     "direction": "better"
    },
    {
     "model": "GPT-6.2 Sol",
     "direction": "better"
    },
    {
     "model": "Gemini 4 Carbon",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X055",
   "voice": "Simon Willison",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/simonw/status/2122844219270422750",
   "quote": "I was most impressed with Fable because it did at least have the physical geography of Yonkers form together into the mech - but none of them are excellent",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X056",
   "voice": "Simon Willison",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/simonw/status/2122844219270422750",
   "quote": "I was most impressed with Fable because it did at least have the physical geography of Yonkers form together into the mech - but none of them are excellent",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X057",
   "voice": "Simon Willison",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/simonw/status/2120258951594037732",
   "quote": "Gemini 4 Carbon is good at pelican SVGs, but that's nothing new for Gemini",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X058",
   "voice": "Simon Willison",
   "model": "Gemini 4 Hydrogen",
   "date": "2026-11-13",
   "platform": "x",
   "url": "https://x.com/simonw/status/2121039423441981444",
   "quote": "The pricing for Gemini 4 is pretty confusing - Argon and Carbon are the same token price as each other, and the new Hydrogen is only a 25% discount on that price - though Carbon/Argon will 2x in price on Dec 1st Hydrogen's pelicans are notably worse",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Gemini 4 Carbon",
     "direction": "worse"
    },
    {
     "model": "Gemini 4 Argon",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X059",
   "voice": "Simon Willison",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-23",
   "platform": "x",
   "url": "https://x.com/simonw/status/2113718487529267593",
   "quote": "More notes on Kimi K3.1 - this one is I big deal, first time an open weights (well, soon-to-be open weights) model has competed this credibly with the current frontier",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X060",
   "voice": "Ethan Mollick",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/emollick/status/2122755693191287009",
   "quote": "I’ve been using GPT-6.2 Astra prior to launch & I found it very uneven & weird. On one hand, it was harder to steer & work with than other models. On the other, when given an ambitious project (make a space elevator ride from scratch with Blender & Unity) it could deliver well",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X061",
   "voice": "Ethan Mollick",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/emollick/status/2122761622468894946",
   "quote": "This is the first time since GPT 4.5 that a frontier release from OpenAi is not a clear step forward. I’m willing to bet that 6.2 is amazing at some types of complex work, but in my experience & that of others I know, it was often frustrating to use.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X062",
   "voice": "Ethan Mollick",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/emollick/status/2122757411287318584",
   "quote": "Fable and opus and carbon all have more ambition, seem to have better anticipation of what your needs are, and generally have better taste on research or innovation topics",
   "basis": "hands-on",
   "domain": "research",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    },
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    },
    {
     "model": "Gemini 4 Carbon",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X063",
   "voice": "Ethan Mollick",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/emollick/status/2122756233291821250",
   "quote": "Though this impressive (& it is), I don’t find myself reaching for the new 6.2 models. They really are hard to work with & take more guidance and pushing to get work done.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X064",
   "voice": "Ethan Mollick",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/emollick/status/2119870150845599972",
   "quote": "I had early access to Fable 5.2 for a bit & I think it crosses the line into being a sort of Jagged AGI.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X065",
   "voice": "Ethan Mollick",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/emollick/status/2120000888974766501",
   "quote": "Looks like Fable 5.2 is above the already steep trendline on three of the most informative unsaturated benchmarks",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "X066",
   "voice": "Ethan Mollick",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/emollick/status/2120198832202125322",
   "quote": "Had early access to Gemini 4 Carbon, it is a very good model that improves on Argon. It is useful to have an AI that thinks differently than Claude/GPT & is still highly capable, but has its own strengths & weakness",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Gemini 4 Argon",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X067",
   "voice": "Ethan Mollick",
   "model": "Kimi K3.1 Pro",
   "date": "2026-11-16",
   "platform": "x",
   "url": "https://x.com/emollick/status/2122323330988671063",
   "quote": "Kimi is good. It is not Fable though. Brittle outside of coding in my experience.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X068",
   "voice": "Ethan Mollick",
   "model": "Qwen4-Max",
   "date": "2026-11-16",
   "platform": "x",
   "url": "https://x.com/emollick/status/2122308260325720127",
   "quote": "I think contrary to many of our expectations, we have not seen a Fable class open model more than 8 months after Fable/Mythos was first made public.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": -1
  },
  {
   "id": "X069",
   "voice": "Matt Shumer",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/mattshumer_/status/2119867949309714609",
   "quote": "Fable 5.2 is the best model ever created. The things I created with it are mind-blowing. It’s also basically a school hall monitor.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X070",
   "voice": "Matt Shumer",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/mattshumer_/status/2119880176112660707",
   "quote": "Btw the new meta is: - Fable 5.2 for planning - Opus 5.5 for execution Trust me, and thank me later.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X071",
   "voice": "Matt Shumer",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/mattshumer_/status/2119884041331892403",
   "quote": "Fable 5.2 ONE-SHOTTED this video. It makes Opus look like a toy.",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X072",
   "voice": "Matt Shumer",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/mattshumer_/status/2119888685709881582",
   "quote": "Fable 5.2 has one fatal flaw: it’s a hall monitor. I charted how often I swear at my agents each day. Guess which day I got access.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X073",
   "voice": "Matt Shumer",
   "model": "Claude Fable 5.2",
   "date": "2026-11-13",
   "platform": "x",
   "url": "https://x.com/mattshumer_/status/2121081410882572721",
   "quote": "This Fable/Opus narc behavior is infuriating",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X074",
   "voice": "Matt Shumer",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/mattshumer_/status/2122752192226406421",
   "quote": "I will not be writing a full review of GPT-6.2 Astra / Sol. I had a terrible experience testing them. They were essentially unusable. I do not recommend these models.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X075",
   "voice": "Matt Shumer",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/mattshumer_/status/2122752192226406421",
   "quote": "I will not be writing a full review of GPT-6.2 Astra / Sol. I had a terrible experience testing them. They were essentially unusable. I do not recommend these models.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X076",
   "voice": "Matt Shumer",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/mattshumer_/status/2122753395509383647",
   "quote": "this is the first launch that is just downright negative. Both models are worse than Opus 5.5",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X077",
   "voice": "Matt Shumer",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/mattshumer_/status/2122753395509383647",
   "quote": "this is the first launch that is just downright negative. Both models are worse than Opus 5.5",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X078",
   "voice": "Matt Shumer",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/mattshumer_/status/2122826912406126786",
   "quote": "FWIW, my opus/fable agents tried it as a sub-agent and rejected most of what it did for them, on the basis that the work sucked and they didn’t follow instructions lol",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    },
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X079",
   "voice": "Matt Shumer",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/mattshumer_/status/2122878504333087108",
   "quote": "I've been testing GPT-6.2 for weeks. It's utterly unusable. Here's my review:",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X080",
   "voice": "Flavio Adamo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/flavioAd/status/2122761415362814076",
   "quote": "6.2 Astra is incredible at 3D, better than Fable or any other model I’ve tried. It’s also less of a yes-man than Astra 6 and incredibly efficient",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X081",
   "voice": "Flavio Adamo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/flavioAd/status/2122761415362814076",
   "quote": "on the frontend side, Astra 6.2 tends to follow its own style. When I ask it to add a new component or section, it often doesn’t match the existing design unless I explicitly tell it to, so I kept going back to Opus.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X082",
   "voice": "Flavio Adamo",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/flavioAd/status/2122987673975398838",
   "quote": "I think Astra 6.2 is a good model. I don’t think it’s better than Fable or Opus, but that doesn’t automatically make it bad",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    },
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 1
  },
  {
   "id": "X083",
   "voice": "Flavio Adamo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/flavioAd/status/2119874593305584103",
   "quote": "Claude Fable 5.2 first impressions: - LOVES using subagents - noticeably faster than Fable 5.1 - oh god, it's SO good at frontend super happy with it so far",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.1",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X084",
   "voice": "Flavio Adamo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/flavioAd/status/2119876761299059127",
   "quote": "WOW Claude Fable 5.2 is insanely good at 3D",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X085",
   "voice": "Flavio Adamo",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/flavioAd/status/2120170537963901288",
   "quote": "My thoughts on Fable 5.2 after 1 day: I love it. It's noticeably better than Opus at figuring things out, although sometimes it thinks way too much about really simple problems, even on medium",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 2
  },
  {
   "id": "X086",
   "voice": "Flavio Adamo",
   "model": "GPT-6.1 Terra",
   "date": "2026-10-26",
   "platform": "x",
   "url": "https://x.com/flavioAd/status/2114791609279631615",
   "quote": "I had early access and used it a lot in ChatGPT Work More of my work lately has involved documents rather than code, so I've really come to appreciate how it writes! (and it's fast)",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X087",
   "voice": "Pietro Schirano",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/skirano/status/2122770640604111133",
   "quote": "I have been early testing Astra 6.2 and I feel it's getting an unfair response. From my testing, most of the issues were with the harness (I even brought it up during testing). Once I removed that, I was flying.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X088",
   "voice": "Pietro Schirano",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/skirano/status/2122772552783376857",
   "quote": "Here's a few one shot things \"a nostalgic arcade racing game set in Japan\" 6.2 even made the music and multiple levels.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "X089",
   "voice": "Pietro Schirano",
   "model": "Claude Fable 5.2",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/skirano/status/2120202483574403350",
   "quote": "Fable 5.2 is such an interesting model. Capable of absolute brilliance, yet incredibly dumb when it comes to the simplest things.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X090",
   "voice": "Pietro Schirano",
   "model": "GPT-6.1 Terra",
   "date": "2026-10-27",
   "platform": "x",
   "url": "https://x.com/skirano/status/2115228389195862341",
   "quote": "People hating on Terra is wild. The model is basically free in Codex.",
   "basis": "hands-on",
   "domain": "price-speed",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X091",
   "voice": "Peter Yang",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/petergyang/status/2123121608860762307",
   "quote": "Here's a use case that GPT-6.2 Astra IS good at: design. I asked Astra, Gemini 4 Carbon, and Fable 5.2 to redesign my website. Astra produced the cleanest version, while Fable had the boldest visual direction and the best motion graphics.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Gemini 4 Carbon",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X092",
   "voice": "Peter Yang",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/petergyang/status/2123121608860762307",
   "quote": "Carbon, in comparison, overstuffed the UI way too much. It felt like it was designing for agents to read instead of humans.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6.2 Astra",
     "direction": "worse"
    },
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X093",
   "voice": "Peter Yang",
   "model": "Claude Fable 5.2",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/petergyang/status/2123121608860762307",
   "quote": "Astra produced the cleanest version, while Fable had the boldest visual direction and the best motion graphics.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Gemini 4 Carbon",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X094",
   "voice": "Peter Yang",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/petergyang/status/2123174863397650705",
   "quote": "Fable and Carbon both did a better job. I prefer Gemini 4 because its passages are more descriptive. So tl:dr I think Gemini is the best model in the market today for writing",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "better"
    },
    {
     "model": "GPT-6.2 Astra",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X095",
   "voice": "Peter Yang",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/petergyang/status/2123174863397650705",
   "quote": "Astra wrote about coffee and a hospital cafeteria instead of the actual topic.",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Gemini 4 Carbon",
     "direction": "worse"
    },
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X096",
   "voice": "Peter Yang",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/petergyang/status/2122927909585432600",
   "quote": "6.2 feels very lazy it doesn't work as hard. Maybe I need to switch from Medium to Hard effort",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X097",
   "voice": "Peter Yang",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/petergyang/status/2119870223562191289",
   "quote": "I had early access and tested @claudeai 's new model head to head against Opus 5.5 across 5 use cases ... Fable won 4 of 5, but Opus still won one.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X098",
   "voice": "Peter Yang",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/petergyang/status/2119874681130103272",
   "quote": "I'm not a SWE but I like using Fable to audit my code base for bugs and issues then getting Opus to fix them. Fable is more thorough than Opus from my testing.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X099",
   "voice": "Peter Yang",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/petergyang/status/2119873615558836510",
   "quote": "Nope Fable is a better writer based on my tests",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X100",
   "voice": "Peter Yang",
   "model": "Gemini 4 Neon",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/petergyang/status/2123189119937601902",
   "quote": "Yeah try neon it’s Gemini equivalent of opus",
   "basis": "announcement",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 1
  },
  {
   "id": "X101",
   "voice": "Matthew Berman",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/MatthewBerman/status/2119871207206158459",
   "quote": "Fable 5.2 is HERE! 🔥 I've burned over 100 billion tokens early testing Fable 5.2. It is BY FAR the best model I've ever used.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X102",
   "voice": "Matthew Berman",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/MatthewBerman/status/2119868627885441389",
   "quote": "Fable 5.2 is absolutely insane. And yes, this is why I say Software is Solved. I've had early access and here are the most incredible demos I created",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X103",
   "voice": "Matthew Berman",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/MatthewBerman/status/2119868633086378471",
   "quote": "Motion graphics are completely solved. I asked Fable 5.2 to create a mograph to explain how LLMs work:",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X104",
   "voice": "Matthew Berman",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/MatthewBerman/status/2122760062326522031",
   "quote": "Been testing these models for a while. They are good, although I’ve seen others really dislike them. They are not as good as Opus 5.5 and Fable. OpenAI came up short.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    },
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X105",
   "voice": "Matthew Berman",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/MatthewBerman/status/2122760062326522031",
   "quote": "Been testing these models for a while. They are good, although I’ve seen others really dislike them. They are not as good as Opus 5.5 and Fable. OpenAI came up short.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    },
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X106",
   "voice": "Matthew Berman",
   "model": "DeepSeek Flash 1017",
   "date": "2026-10-27",
   "platform": "x",
   "url": "https://x.com/MatthewBerman/status/2115129395195302282",
   "quote": "deepseek v4.1 flash is the best open weights model on the planet right now",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X107",
   "voice": "Matthew Berman",
   "model": "DeepSeek Flash 1017",
   "date": "2026-10-27",
   "platform": "x",
   "url": "https://x.com/MatthewBerman/status/2115183971789345118",
   "quote": "I just couldn't believe how good it is. Especially after comparing to Kimi 3.1",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Kimi K3.1 Pro",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X108",
   "voice": "Matthew Berman",
   "model": "Kimi K3.1 Pro",
   "date": "2026-10-24",
   "platform": "x",
   "url": "https://x.com/MatthewBerman/status/2114035962489851904",
   "quote": "Kimi K3.1 smells like Claude",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 0
  },
  {
   "id": "X109",
   "voice": "Matthew Berman",
   "model": "Gemini 4 Carbon",
   "date": "2026-10-28",
   "platform": "x",
   "url": "https://x.com/MatthewBerman/status/2115503276108194165",
   "quote": "Annndddd Google is in the lead.... Absurd gains from Argon to Carbon.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Gemini 4 Argon",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X110",
   "voice": "Matthew Berman",
   "model": "Qwen4-Max",
   "date": "2026-11-16",
   "platform": "x",
   "url": "https://x.com/MatthewBerman/status/2122158687011897726",
   "quote": "Qwen 4 is here! Not open weights unfortunately but a smaller version is coming soon that’ll be open weights.",
   "basis": "announcement",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X111",
   "voice": "ThePrimeagen",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/ThePrimeagen/status/2123259481312108937",
   "quote": "I know I am a bit of a contrarian... I like sol 6.2",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X112",
   "voice": "ThePrimeagen",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/ThePrimeagen/status/2123188274944717013",
   "quote": "\"Hey sol, use my computer and create tokens for cloudflare, youtube, twitch, kick, x and then make sure sign in works locally and on production\" 5 minutes later... I never have to use google console ever again",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X113",
   "voice": "ThePrimeagen",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/ThePrimeagen/status/2123084099581796693",
   "quote": "Astra 6.2 Ultrafast is miles ahead of the competition. Just look at this 2d to 3d transformation it did in 20 short minutes and 4% of my weekly usage",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 0,
   "stance_second": 2
  },
  {
   "id": "X114",
   "voice": "ThePrimeagen",
   "model": "Claude Fable 5.2",
   "date": "2026-11-15",
   "platform": "x",
   "url": "https://x.com/ThePrimeagen/status/2122053127130222881",
   "quote": "All of my work is so trivial that I cannot tell a difference between fable 5.2 and opus 5.5 high",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "same"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X115",
   "voice": "ThePrimeagen",
   "model": "Claude Fable 5.2",
   "date": "2026-11-16",
   "platform": "x",
   "url": "https://x.com/ThePrimeagen/status/2122332661108683033",
   "quote": "I'm sure there's work that actually does work better for fable, and perhaps this is one of those that you're holding it wrong moments for me, but I'm still building software slow and thoughtful and so perhaps I miss out on some of the power But also it makes terrible decisions",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X116",
   "voice": "ThePrimeagen",
   "model": "Muse Ember",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/ThePrimeagen/status/2119937367674937848",
   "quote": "I know the hotness is Fable right now, but Muse Ember is shaping up to be an amazing utility player. Highly recommend it for day to day automation (experience is still limited, hopefully more to post in the next few hours)",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 2
  },
  {
   "id": "X117",
   "voice": "ThePrimeagen",
   "model": "DeepSeek Flash 1017",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/ThePrimeagen/status/2119961602560835686",
   "quote": "for pipeline qa stuff, kind of sucked and only moderate speed up. So meh there * opus 5.5 orchestrator + deepseek implementor / search... holy f, that was awesome. I want that 1ktps all day every day now.",
   "basis": "hands-on",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 0
  },
  {
   "id": "X118",
   "voice": "ThePrimeagen",
   "model": "Gemini 4 Argon",
   "date": "2026-10-31",
   "platform": "x",
   "url": "https://x.com/ThePrimeagen/status/2116591623907766326",
   "quote": "Don't get mad at me Google cooked with Argon 4",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "X119",
   "voice": "Paweł Huryn",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/PawelHuryn/status/2120283296550547618",
   "quote": "What a day! I tested Gemini 4 Carbon on real work. 2 repos, 105 replanted bugs. Find and fix what you can. It scores just behind Fable 5.2.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    },
    {
     "model": "Gemini 4 Argon",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X120",
   "voice": "Paweł Huryn",
   "model": "Gemini 4 Hydrogen",
   "date": "2026-11-13",
   "platform": "x",
   "url": "https://x.com/PawelHuryn/status/2121286545223086302",
   "quote": "The first results for Gemini 4 Hydrogen are in! 2 real repos, 105 replanted bugs, find and fix what you can. - Stronger and cheaper than Haiku 5.5 (max) - Stronger and faster than GPT-6 Luna (max) - Gemini 4 Hydrogen (xhigh) is just behind Sonnet 5.5 (xhigh)",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Haiku 5.5",
     "direction": "better"
    },
    {
     "model": "GPT-6 Luna",
     "direction": "better"
    },
    {
     "model": "Claude Sonnet 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X121",
   "voice": "Paweł Huryn",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/PawelHuryn/status/2123011354843902128",
   "quote": "I tested GPT-6.2 Astra on real work. 2 repos, 105 replanted bugs. Find and fix what you can. It's the new #1. ... Whatever people say about other use cases, it's a code review beast.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "better"
    },
    {
     "model": "Gemini 4 Carbon",
     "direction": "better"
    },
    {
     "model": "GPT-6 Astra",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X122",
   "voice": "Chubby (kimmonismus)",
   "model": "Gemini 4 Hydrogen",
   "date": "2026-11-12",
   "platform": "x",
   "url": "https://x.com/kimmonismus/status/2120940962268770672",
   "quote": "Oh, and here are the benchmarks for Gemini 4 Hydrogen: It handles tasks 2.5× faster than 3.8 Flash and outperforms Argon on the ArtificialAnlys Intelligence Index. Yes, google is on an insane run. I love the model.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Gemini 4 Argon",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X123",
   "voice": "Chubby (kimmonismus)",
   "model": "Gemini 4 Neon",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/kimmonismus/status/2123169294112698774",
   "quote": "The benchmarks for Gemini 4 Neon are quite impressive. I wasn’t aware of these benchmarks beforehand, either. In just a few months, Google has managed to turn things around and is back in the game.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "X124",
   "voice": "Chubby (kimmonismus)",
   "model": "Gemini 4 Neon",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/kimmonismus/status/2123221275594403932",
   "quote": "Gemini 4 Neon even surpassed Fable 5.2 in coding agent index and sits 6 points ahead of the yesterday released GPT-6.2 Astra.",
   "basis": "benchmark-reaction",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "better"
    },
    {
     "model": "GPT-6.2 Astra",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "X125",
   "voice": "Chubby (kimmonismus)",
   "model": "Qwen4-Max",
   "date": "2026-11-16",
   "platform": "x",
   "url": "https://x.com/kimmonismus/status/2122294330778132787",
   "quote": "Qwen 4 is not a good release. Although it appears on paper to be on par with Opus 5.5 in key benchmarks, the AA Index paints a completely different picture.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -1
  },
  {
   "id": "X126",
   "voice": "McKay Wrigley",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/mckaywrigley/status/2122777818434265318",
   "quote": "astra 6.2 is a pretty good ad for anthropic’s ipo",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -1
  },
  {
   "id": "X127",
   "voice": "McKay Wrigley",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/mckaywrigley/status/2122780014437634266",
   "quote": "maybe it benchmarks better, but it feels like a regression in every way. initial gpt astra release had a lot of magic that i feel like subsequent updates have lacked.",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6 Astra",
     "direction": "worse"
    }
   ],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X128",
   "voice": "McKay Wrigley",
   "model": "GPT-6.2 Sol",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/mckaywrigley/status/2122799598892912749",
   "quote": "at least sol is good!",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X129",
   "voice": "Riley Brown",
   "model": "Claude Fable 5.2",
   "date": "2026-11-11",
   "platform": "x",
   "url": "https://x.com/rileybrown/status/2120385367618486370",
   "quote": "Fable 5.2 is the best model ever released.. and it's not close. Especially at any visual task.",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X130",
   "voice": "Riley Brown",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/rileybrown/status/2122812863492317294",
   "quote": "i don't have any thoughts on the new models released by OpenAI other than I will be using Opus 5.5 for about 85% of things and Fable for 15% of things until someone else catches up.",
   "basis": "hearsay",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    },
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X131",
   "voice": "Deedy Das",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/deedydas/status/2119930300990177509",
   "quote": "Wow, Fable 5.2 is unbelievable at generating photorealistic scenes.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "investor",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X132",
   "voice": "Deedy Das",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/deedydas/status/2122818971854639244",
   "quote": "OpenAI tried to launch a new model before the big day but GPT 6.2 Astra has been a massive fail for every critic that has reviewed it.",
   "basis": "hearsay",
   "domain": "general",
   "disclosure": "investor",
   "compared_to": [],
   "stance_first": -2,
   "stance_second": -2
  },
  {
   "id": "X133",
   "voice": "Deedy Das",
   "model": "Mercury 3",
   "date": "2026-10-20",
   "platform": "x",
   "url": "https://x.com/deedydas/status/2112447082883899709",
   "quote": "Inception just dropped Mercury 3, a model 4-8x faster than Luna / Haiku for the same price and benchmark performance. In real world use, OpenRouter shows that it’s effectively 2-4x faster and 2-4x cheaper effective price.",
   "basis": "benchmark-reaction",
   "domain": "price-speed",
   "disclosure": "investor",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 1
  },
  {
   "id": "X134",
   "voice": "Zvi Mowshowitz",
   "model": "Claude Fable 5.2",
   "date": "2026-11-12",
   "platform": "x",
   "url": "https://x.com/TheZvi/status/2120847873151549877",
   "quote": "Fable 5.2 is a clear downgrade on EditorBench versus Opus 5.5, reverting.",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X135",
   "voice": "Zvi Mowshowitz",
   "model": "Claude Fable 5.2",
   "date": "2026-11-12",
   "platform": "x",
   "url": "https://x.com/TheZvi/status/2120988684002500987",
   "quote": "Fable 5.2 has encouraged me to expand my ambitions once again and more aggressively code up things, except that Opus 5.5 so far has done everything first so I haven't actually had to use Fable 5.2 for that yet.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "same"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X136",
   "voice": "Kieran Klaassen (Every)",
   "model": "Claude Fable 5.2",
   "date": "2026-11-09",
   "platform": "x",
   "url": "https://x.com/kieranklaassen/status/2119887993674821762",
   "quote": "Fable 5.2 is really good at coding. It leaved the most maintainable code of any model. The code you want to keep around for many years. I tested it against all other models for a long time and really think it's great.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X137",
   "voice": "Kieran Klaassen (Every)",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-11",
   "platform": "x",
   "url": "https://x.com/kieranklaassen/status/2120450240507523249",
   "quote": "I really like Carbon for coding. Still behind Opus and Fable, but I'm using it for stuff... :)",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "worse"
    },
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X138",
   "voice": "Kieran Klaassen (Every)",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/kieranklaassen/status/2123210240481186034",
   "quote": "Coding with Astra all day. Do not like it a bit being in the loop.",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X139",
   "voice": "Kieran Klaassen (Every)",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/kieranklaassen/status/2122775824961298484",
   "quote": "GTT-6 Astra 6.2 LOVES mountains. Mountains is the new slop. All non related separate runs that never mention mountains ever:",
   "basis": "hands-on",
   "domain": "writing",
   "disclosure": "early-access",
   "compared_to": [],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X140",
   "voice": "Chetaslua",
   "model": "Gemini 4 Neon",
   "date": "2026-11-19",
   "platform": "x",
   "url": "https://x.com/chetaslua/status/2123219124260315301",
   "quote": "Gemini 4 Neon made a motion design to introduce herself. Pure JS code, not a single external asset.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X141",
   "voice": "Chetaslua",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/chetaslua/status/2123072164895277284",
   "quote": "Gemini 4 Carbon is the best model in the world for me. It sounds original and thinks in a way no other model does",
   "basis": "hands-on",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X142",
   "voice": "Chetaslua",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-15",
   "platform": "x",
   "url": "https://x.com/chetaslua/status/2121885449509851415",
   "quote": "Gemini 4 Carbon with Hydrogen swarm < best thing ever you can do with Antigravity > This art piece is pure code",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X143",
   "voice": "Chetaslua",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/chetaslua/status/2122752021220716975",
   "quote": "Holy Shit this is generational leap Astra 6 vs Astra 6.2 , this is the spatial reasoning i was talking about for the last 1 month @OpenAIDevs you guys cooked , thanks for this model",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6 Astra",
     "direction": "better"
    }
   ],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X144",
   "voice": "Chetaslua",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/chetaslua/status/2122779658815193467",
   "quote": "GPT-6.2 Astra : got 2x worse at telling you when it's stuck > hides a blocker (missing file / dead tool) 22.3% of the time. GPT-6 Astra was 11.1% its so over for openai",
   "basis": "benchmark-reaction",
   "domain": "agentic",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-6 Astra",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -2
  },
  {
   "id": "X145",
   "voice": "Chetaslua",
   "model": "Claude Fable 5.2",
   "date": "2026-11-13",
   "platform": "x",
   "url": "https://x.com/chetaslua/status/2121176151234330707",
   "quote": "Fable 5.2 can achieve this with @threejs < no external assets at all>",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X146",
   "voice": "Pankaj Kumar",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/pankajkumar_dev/status/2122785662734868949",
   "quote": "First output of GPT-6.2 Astra from my side! Its actually quite good. I would say Fable level, and it is really strong at 3D.",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "same"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X147",
   "voice": "Pankaj Kumar",
   "model": "GPT-6.2 Astra",
   "date": "2026-11-17",
   "platform": "x",
   "url": "https://x.com/pankajkumar_dev/status/2122751520286638140",
   "quote": "GPT-6.2 Astra scores 62.1 on the AA Intelligence Index AND GPT-6.2 Sol scores 57.3. tldr : Anthropic is better.",
   "basis": "benchmark-reaction",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Fable 5.2",
     "direction": "worse"
    }
   ],
   "stance_first": -1,
   "stance_second": -1
  },
  {
   "id": "X148",
   "voice": "Pankaj Kumar",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-11",
   "platform": "x",
   "url": "https://x.com/pankajkumar_dev/status/2120372310691742043",
   "quote": "Gemini 4 Carbon : My Impressions/Review Pros: - Its a good jump over Argon. Its quite good at understanding videos, audio and images, which Argon also had, but Carbon does it much better. - It has very good taste.",
   "basis": "hands-on",
   "domain": "multimodal",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Gemini 4 Argon",
     "direction": "better"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X149",
   "voice": "Pankaj Kumar",
   "model": "Gemini 4 Carbon",
   "date": "2026-11-10",
   "platform": "x",
   "url": "https://x.com/pankajkumar_dev/status/2120203148883366162",
   "quote": "This is very close to Opus 5.5 in terms of visual quality, but Carbon has its own design taste.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Claude Opus 5.5",
     "direction": "same"
    }
   ],
   "stance_first": 1,
   "stance_second": 1
  },
  {
   "id": "X150",
   "voice": "Pankaj Kumar",
   "model": "Gemini 4 Hydrogen",
   "date": "2026-11-13",
   "platform": "x",
   "url": "https://x.com/pankajkumar_dev/status/2121220537984954448",
   "quote": "Gemini 4 Hydrogen is f*ckn good, man. I asked it to generate a 3D miniature Post Office on a floating island. Just look at the textures, man. Almost looks real.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X151",
   "voice": "Pankaj Kumar",
   "model": "Claude Fable 5.2",
   "date": "2026-11-11",
   "platform": "x",
   "url": "https://x.com/pankajkumar_dev/status/2120524447383683270",
   "quote": "Fable 5.2 is very good at building games. I asked it to build a 3D Tokyo drifting game with AAA-style visuals. And it actually built the whole game, and it genuinely has that AAA vibe.",
   "basis": "hands-on",
   "domain": "frontend-design",
   "disclosure": "none",
   "compared_to": [],
   "stance_first": 2,
   "stance_second": 2
  },
  {
   "id": "X152",
   "voice": "Pankaj Kumar",
   "model": "Gemini 4 Neon",
   "date": "2026-11-18",
   "platform": "x",
   "url": "https://x.com/pankajkumar_dev/status/2123133228915785958",
   "quote": "Gemini 4 Neon is here! It's the Google's fastest reasoning model and comes close to Carbon, but doesn't exceed it.",
   "basis": "announcement",
   "domain": "general",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "Gemini 4 Carbon",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  },
  {
   "id": "X153",
   "voice": "Paweł Huryn",
   "model": "GPT-6.1 Terra",
   "date": "2026-10-26",
   "platform": "x",
   "url": "https://x.com/PawelHuryn/status/2114860186330398962",
   "quote": "It's cost and time effective. But right after max, its performance drops dramatically - far below Haiku 5.5 (xhigh).",
   "basis": "hands-on",
   "domain": "coding",
   "disclosure": "none",
   "compared_to": [
    {
     "model": "GPT-5.6 Terra",
     "direction": "better"
    },
    {
     "model": "Claude Haiku 5.5",
     "direction": "worse"
    }
   ],
   "stance_first": 0,
   "stance_second": 0
  }
 ],
 "vibe_score": {
  "Gemini 4 Neon": {
   "score": 1.38,
   "voices": 4,
   "records": 7
  },
  "Muse Ember": {
   "score": 1.25,
   "voices": 4,
   "records": 4
  },
  "DeepSeek-V4.1-Flash-1017": {
   "score": 1.0,
   "voices": 3,
   "records": 7
  },
  "Claude Fable 5.2": {
   "score": 0.88,
   "voices": 21,
   "records": 75
  },
  "Gemini 4 Carbon": {
   "score": 0.85,
   "voices": 14,
   "records": 40
  },
  "GPT-6.1 Terra": {
   "score": 0.73,
   "voices": 5,
   "records": 7
  },
  "Gemini 4 Argon": {
   "score": 0.33,
   "voices": 3,
   "records": 3
  },
  "Gemini 4 Hydrogen": {
   "score": 0.15,
   "voices": 6,
   "records": 10
  },
  "Kimi K3.1 Pro": {
   "score": 0.08,
   "voices": 7,
   "records": 12
  },
  "Qwen4-Max": {
   "score": 0.04,
   "voices": 3,
   "records": 7
  },
  "GPT-6.2 Astra": {
   "score": -0.01,
   "voices": 17,
   "records": 57
  },
  "GPT-6.2 Sol": {
   "score": -0.06,
   "voices": 9,
   "records": 13
  }
 },
 "rank_correlation_with_vibe": [
  {
   "benchmark": "LiveBench 2.0 design",
   "n": 9,
   "rho": 0.17,
   "ci90": [
    0.03,
    0.62
   ]
  },
  {
   "benchmark": "DesignArena Overall Agentic",
   "n": 8,
   "rho": 0.08,
   "ci90": [
    -0.24,
    0.4
   ]
  },
  {
   "benchmark": "LiveBench 2.0 long_agentic",
   "n": 9,
   "rho": 0.05,
   "ci90": [
    -0.13,
    0.57
   ]
  },
  {
   "benchmark": "Vals Index",
   "n": 11,
   "rho": 0.05,
   "ci90": [
    -0.19,
    0.26
   ]
  },
  {
   "benchmark": "Arena Agent (Agent Arena Index)",
   "n": 11,
   "rho": 0.04,
   "ci90": [
    -0.14,
    0.31
   ]
  },
  {
   "benchmark": "BulletBench Blitz 3+2",
   "n": 8,
   "rho": 0.02,
   "ci90": [
    -0.36,
    0.31
   ]
  },
  {
   "benchmark": "LiveBench 2.0 coding",
   "n": 9,
   "rho": -0.02,
   "ci90": [
    -0.13,
    0.53
   ]
  },
  {
   "benchmark": "CatalogBench (factual brief)",
   "n": 11,
   "rho": -0.04,
   "ci90": [
    -0.25,
    0.21
   ]
  },
  {
   "benchmark": "LiveBench 2.0 overall",
   "n": 9,
   "rho": -0.07,
   "ci90": [
    -0.18,
    0.48
   ]
  },
  {
   "benchmark": "AA Intelligence Index",
   "n": 11,
   "rho": -0.11,
   "ci90": [
    -0.26,
    0.2
   ]
  },
  {
   "benchmark": "LiveBench 2.0 research",
   "n": 9,
   "rho": -0.12,
   "ci90": [
    -0.32,
    0.37
   ]
  },
  {
   "benchmark": "LiveBench 2.0 data_analysis",
   "n": 9,
   "rho": -0.13,
   "ci90": [
    -0.39,
    0.38
   ]
  },
  {
   "benchmark": "CatalogBench (factual brief) [reliable]",
   "n": 10,
   "rho": -0.15,
   "ci90": [
    -0.36,
    0.18
   ]
  },
  {
   "benchmark": "ROASBench",
   "n": 11,
   "rho": -0.22,
   "ci90": [
    -0.37,
    0.16
   ]
  },
  {
   "benchmark": "AA Terminal-Bench 4.0 (in-index run)",
   "n": 6,
   "rho": -0.26,
   "ci90": [
    -0.6,
    0.03
   ]
  },
  {
   "benchmark": "SurveyBench",
   "n": 11,
   "rho": -0.32,
   "ci90": [
    -0.41,
    0.02
   ]
  },
  {
   "benchmark": "Overall leaderboard (Spring Prompt, blends third-party sources)",
   "n": 11,
   "rho": -0.33,
   "ci90": [
    -0.44,
    0.01
   ]
  },
  {
   "benchmark": "Spring Prompt Score",
   "n": 11,
   "rho": -0.35,
   "ci90": [
    -0.47,
    0.03
   ]
  },
  {
   "benchmark": "Arena WebDev",
   "n": 11,
   "rho": -0.37,
   "ci90": [
    -0.47,
    -0.05
   ]
  },
  {
   "benchmark": "OpenRouter token share, week 9-15 Nov",
   "n": 6,
   "rho": -0.37,
   "ci90": [
    -0.94,
    0.14
   ]
  },
  {
   "benchmark": "Terminal-Bench 4.1 (tbench.ai)",
   "n": 9,
   "rho": -0.38,
   "ci90": [
    -0.52,
    -0.02
   ]
  },
  {
   "benchmark": "CatalogBench (sales brief) [reliable]",
   "n": 10,
   "rho": -0.41,
   "ci90": [
    -0.52,
    0.03
   ]
  },
  {
   "benchmark": "LiveBench 2.0 browser",
   "n": 9,
   "rho": -0.42,
   "ci90": [
    -0.47,
    0.24
   ]
  },
  {
   "benchmark": "Epoch Capabilities Index (ECI)",
   "n": 6,
   "rho": -0.43,
   "ci90": [
    -0.6,
    0.26
   ]
  },
  {
   "benchmark": "BulletBench Bullet 60s",
   "n": 6,
   "rho": -0.46,
   "ci90": [
    -0.82,
    -0.33
   ]
  },
  {
   "benchmark": "CatalogBench (sales brief)",
   "n": 11,
   "rho": -0.48,
   "ci90": [
    -0.6,
    -0.09
   ]
  },
  {
   "benchmark": "DeckBench",
   "n": 11,
   "rho": -0.55,
   "ci90": [
    -0.67,
    -0.21
   ]
  },
  {
   "benchmark": "LiveBench 2.0 reasoning",
   "n": 9,
   "rho": -0.55,
   "ci90": [
    -0.7,
    -0.12
   ]
  }
 ],
 "head_to_head_pairs": {
  "Claude Opus 5.5 vs GPT-6.2 Astra": [
   6,
   0
  ],
  "Claude Fable 5.2 vs GPT-6.2 Astra": [
   8,
   1
  ],
  "Claude Fable 5.2 vs Claude Opus 5.5": [
   8,
   3
  ],
  "Claude Fable 5.2 vs Claude Sonnet 5.5": [
   0,
   2
  ],
  "Claude Opus 5.5 vs Gemini 4 Carbon": [
   2,
   0
  ],
  "Gemini 4 Argon vs Gemini 4 Carbon": [
   0,
   6
  ],
  "Claude Fable 5.2 vs Gemini 4 Carbon": [
   7,
   0
  ],
  "Claude Opus 5.5 vs Qwen4-Max": [
   1,
   0
  ],
  "GPT-6.1 Terra vs Qwen4-Max": [
   1,
   0
  ],
  "Gemini 4 Hydrogen vs Qwen4-Max": [
   1,
   0
  ],
  "Claude Sonnet 5.5 vs Gemini 4 Hydrogen": [
   2,
   0
  ],
  "GPT-6 Astra vs Kimi K3.1 Pro": [
   1,
   0
  ],
  "GPT-6 Astra vs GPT-6.2 Astra": [
   3,
   2
  ],
  "GPT-6.2 Astra vs Gemini 4 Carbon": [
   1,
   3
  ],
  "Claude Fable 5.2 vs GPT-6.2 Sol": [
   3,
   0
  ],
  "GPT-6.2 Sol vs Gemini 4 Carbon": [
   0,
   1
  ],
  "Claude Opus 5.5 vs DeepSeek-V4.1-Flash-1017": [
   1,
   0
  ],
  "Claude Opus 5.5 vs Kimi K3.1 Pro": [
   2,
   0
  ],
  "Claude Haiku 5.5 vs Kimi K3.1 Pro": [
   1,
   0
  ],
  "DeepSeek-V4.1-Flash-1017 vs Kimi K3.1 Pro": [
   2,
   0
  ],
  "Kimi K3.1 Pro vs Qwen4-Max": [
   1,
   0
  ],
  "DeepSeek-V4.1-Flash-1017 vs Qwen4-Max": [
   1,
   0
  ],
  "Claude Fable 5.2 vs GPT-6 Astra": [
   1,
   0
  ],
  "Claude Fable 5.2 vs Gemini 4 Neon": [
   1,
   0
  ],
  "Gemini 4 Carbon vs Gemini 4 Neon": [
   0,
   1
  ],
  "GPT-6.2 Astra vs Muse Wonder Full": [
   0,
   1
  ],
  "GPT-6.1 Sol vs Gemini 4 Hydrogen": [
   1,
   0
  ],
  "Claude Opus 5.5 vs Gemini 4 Hydrogen": [
   1,
   0
  ],
  "Gemini 4 Carbon vs Gemini 4 Hydrogen": [
   2,
   0
  ],
  "Gemini 4 Argon vs Gemini 4 Hydrogen": [
   1,
   0
  ],
  "Claude Fable 5.2 vs Kimi K3.1 Pro": [
   1,
   0
  ],
  "Claude Opus 5.5 vs GPT-6.2 Sol": [
   2,
   0
  ],
  "Claude Fable 5.1 vs Claude Fable 5.2": [
   0,
   1
  ],
  "Claude Haiku 5.5 vs Gemini 4 Hydrogen": [
   0,
   1
  ],
  "GPT-6 Luna vs Gemini 4 Hydrogen": [
   0,
   1
  ],
  "GPT-5.6 Terra vs GPT-6.1 Terra": [
   0,
   1
  ],
  "Claude Haiku 5.5 vs GPT-6.1 Terra": [
   1,
   0
  ]
 },
 "head_to_head_agreement": [
  {
   "benchmark": "BulletBench Blitz 3+2",
   "share_of_votes": 0.77,
   "votes": 30,
   "pairs_agreed": "7/10"
  },
  {
   "benchmark": "CatalogBench (factual brief) [reliable]",
   "share_of_votes": 0.39,
   "votes": 31,
   "pairs_agreed": "4/11"
  },
  {
   "benchmark": "CatalogBench (factual brief)",
   "share_of_votes": 0.69,
   "votes": 35,
   "pairs_agreed": "8/14"
  },
  {
   "benchmark": "CatalogBench (sales brief) [reliable]",
   "share_of_votes": 0.42,
   "votes": 31,
   "pairs_agreed": "5/11"
  },
  {
   "benchmark": "CatalogBench (sales brief)",
   "share_of_votes": 0.4,
   "votes": 35,
   "pairs_agreed": "6/14"
  },
  {
   "benchmark": "DeckBench",
   "share_of_votes": 0.63,
   "votes": 35,
   "pairs_agreed": "6/14"
  },
  {
   "benchmark": "Overall leaderboard (Spring Prompt, blends third-party sources)",
   "share_of_votes": 0.71,
   "votes": 35,
   "pairs_agreed": "9/14"
  },
  {
   "benchmark": "ROASBench",
   "share_of_votes": 0.71,
   "votes": 35,
   "pairs_agreed": "8/14"
  },
  {
   "benchmark": "Spring Prompt Score",
   "share_of_votes": 0.69,
   "votes": 35,
   "pairs_agreed": "8/14"
  },
  {
   "benchmark": "SurveyBench",
   "share_of_votes": 0.66,
   "votes": 35,
   "pairs_agreed": "6/14"
  },
  {
   "benchmark": "AA Intelligence Index",
   "share_of_votes": 0.8,
   "votes": 35,
   "pairs_agreed": "11/14"
  },
  {
   "benchmark": "AA Terminal-Bench 4.0 (in-index run)",
   "share_of_votes": 0.33,
   "votes": 12,
   "pairs_agreed": "3/5"
  },
  {
   "benchmark": "Arena Agent (Agent Arena Index)",
   "share_of_votes": 0.83,
   "votes": 35,
   "pairs_agreed": "11/14"
  },
  {
   "benchmark": "Arena WebDev",
   "share_of_votes": 0.66,
   "votes": 35,
   "pairs_agreed": "6/14"
  },
  {
   "benchmark": "ARC-AGI-2",
   "share_of_votes": 0.25,
   "votes": 24,
   "pairs_agreed": "2/5"
  },
  {
   "benchmark": "ARC-AGI-3 (Provider adapter)",
   "share_of_votes": 0.12,
   "votes": 24,
   "pairs_agreed": "1/5"
  },
  {
   "benchmark": "ARC-AGI-3 (Standard)",
   "share_of_votes": 0.83,
   "votes": 24,
   "pairs_agreed": "4/5"
  },
  {
   "benchmark": "DesignArena Overall Agentic",
   "share_of_votes": 0.72,
   "votes": 29,
   "pairs_agreed": "5/10"
  },
  {
   "benchmark": "Epoch Capabilities Index (ECI)",
   "share_of_votes": 0.43,
   "votes": 23,
   "pairs_agreed": "2/5"
  },
  {
   "benchmark": "LiveBench 2.0 browser",
   "share_of_votes": 0.39,
   "votes": 33,
   "pairs_agreed": "4/12"
  },
  {
   "benchmark": "LiveBench 2.0 coding",
   "share_of_votes": 0.33,
   "votes": 33,
   "pairs_agreed": "7/12"
  },
  {
   "benchmark": "LiveBench 2.0 data_analysis",
   "share_of_votes": 0.39,
   "votes": 33,
   "pairs_agreed": "6/12"
  },
  {
   "benchmark": "LiveBench 2.0 design",
   "share_of_votes": 0.82,
   "votes": 33,
   "pairs_agreed": "9/12"
  },
  {
   "benchmark": "LiveBench 2.0 long_agentic",
   "share_of_votes": 0.88,
   "votes": 33,
   "pairs_agreed": "10/12"
  },
  {
   "benchmark": "LiveBench 2.0 overall",
   "share_of_votes": 0.79,
   "votes": 24,
   "pairs_agreed": "8/11"
  },
  {
   "benchmark": "LiveBench 2.0 reasoning",
   "share_of_votes": 0.12,
   "votes": 32,
   "pairs_agreed": "1/11"
  },
  {
   "benchmark": "LiveBench 2.0 research",
   "share_of_votes": 0.79,
   "votes": 33,
   "pairs_agreed": "8/12"
  },
  {
   "benchmark": "Terminal-Bench 4.1 (tbench.ai)",
   "share_of_votes": 0.86,
   "votes": 29,
   "pairs_agreed": "7/8"
  },
  {
   "benchmark": "Vals Index",
   "share_of_votes": 0.8,
   "votes": 35,
   "pairs_agreed": "10/14"
  },
  {
   "benchmark": "Vending-Bench 2",
   "share_of_votes": 0.27,
   "votes": 11,
   "pairs_agreed": "2/4"
  }
 ],
 "by_reviewer": [
  {
   "voice": "Bijan Bowen",
   "records": 32,
   "models": 10,
   "early_access_share": 0.09,
   "domains": {
    "frontend-design": 13,
    "multimodal": 5,
    "general": 5,
    "price-speed": 4,
    "agentic": 2,
    "coding": 2,
    "writing": 1
   },
   "scores": {
    "Gemini 4 Carbon": 1.5,
    "Hy4": 1.5,
    "Muse Ember": 1.5,
    "Gemini 4 Hydrogen": 0.9,
    "GPT-6.2 Astra": 0.8,
    "DeepSeek-V4.1-Flash-1017": 0.75,
    "GPT-6.1 Terra": 0.67,
    "Claude Fable 5.2": 0.0,
    "Qwen4-Max": -0.12,
    "Kimi K3.1 Pro": -0.5
   },
   "rank_correlation": [
    {
     "benchmark": "LiveBench 2.0 coding",
     "rho": 0.69,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 data_analysis",
     "rho": 0.64,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 long_agentic",
     "rho": 0.52,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 overall",
     "rho": 0.44,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 design",
     "rho": 0.33,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 research",
     "rho": 0.31,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "BulletBench Blitz 3+2",
     "rho": 0.09,
     "n": 6,
     "spring_prompt": true
    },
    {
     "benchmark": "Spring Prompt Score",
     "rho": 0.05,
     "n": 9,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (sales brief) [reliable]",
     "rho": 0.02,
     "n": 8,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 browser",
     "rho": 0.0,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena Agent (Agent Arena Index)",
     "rho": -0.01,
     "n": 10,
     "spring_prompt": false
    },
    {
     "benchmark": "CatalogBench (factual brief) [reliable]",
     "rho": -0.04,
     "n": 8,
     "spring_prompt": true
    },
    {
     "benchmark": "SurveyBench",
     "rho": -0.1,
     "n": 9,
     "spring_prompt": true
    },
    {
     "benchmark": "AA Intelligence Index",
     "rho": -0.13,
     "n": 10,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena WebDev",
     "rho": -0.15,
     "n": 9,
     "spring_prompt": false
    },
    {
     "benchmark": "Vals Index",
     "rho": -0.21,
     "n": 10,
     "spring_prompt": false
    },
    {
     "benchmark": "CatalogBench (sales brief)",
     "rho": -0.24,
     "n": 9,
     "spring_prompt": true
    },
    {
     "benchmark": "ROASBench",
     "rho": -0.28,
     "n": 9,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (factual brief)",
     "rho": -0.28,
     "n": 9,
     "spring_prompt": true
    },
    {
     "benchmark": "Overall leaderboard (Spring Prompt, blends third-party sources)",
     "rho": -0.29,
     "n": 9,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 reasoning",
     "rho": -0.33,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "DesignArena Overall Agentic",
     "rho": -0.35,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "Terminal-Bench 4.1 (tbench.ai)",
     "rho": -0.41,
     "n": 9,
     "spring_prompt": false
    },
    {
     "benchmark": "AA Terminal-Bench 4.0 (in-index run)",
     "rho": -0.52,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "DeckBench",
     "rho": -0.76,
     "n": 9,
     "spring_prompt": true
    }
   ]
  },
  {
   "voice": "Theo",
   "records": 42,
   "models": 9,
   "early_access_share": 0.43,
   "domains": {
    "general": 14,
    "coding": 11,
    "frontend-design": 6,
    "price-speed": 5,
    "agentic": 3,
    "writing": 3
   },
   "scores": {
    "Gemini 4 Neon": 2.0,
    "GPT-6.2 Sol": 1.0,
    "Muse Ember": 1.0,
    "GPT-6.1 Terra": 1.0,
    "Claude Fable 5.2": 0.43,
    "GPT-6.2 Astra": -0.7,
    "Gemini 4 Carbon": -1.38,
    "Gemini 4 Hydrogen": -2.0,
    "Gemini 4 Argon": -2.0
   },
   "rank_correlation": [
    {
     "benchmark": "LiveBench 2.0 browser",
     "rho": 0.78,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "CatalogBench (factual brief) [reliable]",
     "rho": 0.52,
     "n": 8,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (factual brief)",
     "rho": 0.48,
     "n": 8,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (sales brief) [reliable]",
     "rho": 0.37,
     "n": 8,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (sales brief)",
     "rho": 0.37,
     "n": 8,
     "spring_prompt": true
    },
    {
     "benchmark": "Overall leaderboard (Spring Prompt, blends third-party sources)",
     "rho": 0.22,
     "n": 8,
     "spring_prompt": true
    },
    {
     "benchmark": "DeckBench",
     "rho": 0.2,
     "n": 8,
     "spring_prompt": true
    },
    {
     "benchmark": "ROASBench",
     "rho": 0.17,
     "n": 8,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 reasoning",
     "rho": 0.13,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 design",
     "rho": 0.12,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 coding",
     "rho": 0.12,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 overall",
     "rho": 0.09,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "Spring Prompt Score",
     "rho": -0.07,
     "n": 8,
     "spring_prompt": true
    },
    {
     "benchmark": "SurveyBench",
     "rho": -0.1,
     "n": 8,
     "spring_prompt": true
    },
    {
     "benchmark": "DesignArena Overall Agentic",
     "rho": -0.1,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "AA Intelligence Index",
     "rho": -0.15,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena Agent (Agent Arena Index)",
     "rho": -0.17,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 research",
     "rho": -0.2,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 long_agentic",
     "rho": -0.2,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "Vals Index",
     "rho": -0.22,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena WebDev",
     "rho": -0.24,
     "n": 8,
     "spring_prompt": false
    },
    {
     "benchmark": "BulletBench Blitz 3+2",
     "rho": -0.25,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "Terminal-Bench 4.1 (tbench.ai)",
     "rho": -0.44,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 data_analysis",
     "rho": -0.72,
     "n": 6,
     "spring_prompt": false
    }
   ]
  },
  {
   "voice": "Simon Willison",
   "records": 12,
   "models": 8,
   "early_access_share": 0.0,
   "domains": {
    "frontend-design": 10,
    "multimodal": 1,
    "general": 1
   },
   "scores": {
    "Kimi K3.1 Pro": 2.0,
    "Claude Fable 5.2": 1.0,
    "Gemini 4 Carbon": 1.0,
    "Muse Ember": 1.0,
    "Gemini Decision 1": 0.0,
    "GPT-6.2 Astra": -1.0,
    "GPT-6.2 Sol": -1.0,
    "Gemini 4 Hydrogen": -1.0
   },
   "rank_correlation": [
    {
     "benchmark": "AA Terminal-Bench 4.0 (in-index run)",
     "rho": 0.67,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "DesignArena Overall Agentic",
     "rho": 0.63,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "ROASBench",
     "rho": 0.19,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 reasoning",
     "rho": 0.14,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "DeckBench",
     "rho": 0.08,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "Terminal-Bench 4.1 (tbench.ai)",
     "rho": 0.04,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 research",
     "rho": 0.0,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 long_agentic",
     "rho": 0.0,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "BulletBench Blitz 3+2",
     "rho": 0.0,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "Overall leaderboard (Spring Prompt, blends third-party sources)",
     "rho": -0.12,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (factual brief)",
     "rho": -0.14,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "Vals Index",
     "rho": -0.15,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena Agent (Agent Arena Index)",
     "rho": -0.15,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 design",
     "rho": -0.15,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "CatalogBench (sales brief)",
     "rho": -0.18,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 overall",
     "rho": -0.23,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "AA Intelligence Index",
     "rho": -0.27,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 data_analysis",
     "rho": -0.31,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena WebDev",
     "rho": -0.39,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "SurveyBench",
     "rho": -0.46,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 coding",
     "rho": -0.46,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "CatalogBench (sales brief) [reliable]",
     "rho": -0.49,
     "n": 6,
     "spring_prompt": true
    },
    {
     "benchmark": "Spring Prompt Score",
     "rho": -0.58,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 browser",
     "rho": -0.62,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "CatalogBench (factual brief) [reliable]",
     "rho": -0.69,
     "n": 6,
     "spring_prompt": true
    }
   ]
  },
  {
   "voice": "AICodeKing",
   "records": 10,
   "models": 7,
   "early_access_share": 0.0,
   "domains": {
    "coding": 6,
    "price-speed": 2,
    "agentic": 2
   },
   "scores": {
    "Gemini 4 Carbon": 0.5,
    "GPT-6.2 Sol": 0.5,
    "Claude Fable 5.2": 0.0,
    "Gemini 4 Hydrogen": 0.0,
    "Qwen4-Max": 0.0,
    "Kimi K3.1 Pro": -0.75,
    "GPT-6.2 Astra": -1.25
   },
   "rank_correlation": [
    {
     "benchmark": "DesignArena Overall Agentic",
     "rho": 0.39,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "Vals Index",
     "rho": 0.32,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 research",
     "rho": 0.3,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 long_agentic",
     "rho": 0.3,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena Agent (Agent Arena Index)",
     "rho": 0.21,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "Terminal-Bench 4.1 (tbench.ai)",
     "rho": 0.18,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 coding",
     "rho": 0.17,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 design",
     "rho": 0.09,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 data_analysis",
     "rho": 0.09,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena WebDev",
     "rho": 0.09,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 overall",
     "rho": 0.04,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "AA Intelligence Index",
     "rho": 0.02,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 browser",
     "rho": -0.02,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "CatalogBench (sales brief) [reliable]",
     "rho": -0.03,
     "n": 6,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (factual brief) [reliable]",
     "rho": -0.03,
     "n": 6,
     "spring_prompt": true
    },
    {
     "benchmark": "DeckBench",
     "rho": -0.07,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "Spring Prompt Score",
     "rho": -0.09,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (factual brief)",
     "rho": -0.09,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "SurveyBench",
     "rho": -0.19,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (sales brief)",
     "rho": -0.19,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "AA Terminal-Bench 4.0 (in-index run)",
     "rho": -0.22,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "BulletBench Blitz 3+2",
     "rho": -0.37,
     "n": 6,
     "spring_prompt": true
    },
    {
     "benchmark": "ROASBench",
     "rho": -0.39,
     "n": 7,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 reasoning",
     "rho": -0.4,
     "n": 7,
     "spring_prompt": false
    },
    {
     "benchmark": "Overall leaderboard (Spring Prompt, blends third-party sources)",
     "rho": -0.41,
     "n": 7,
     "spring_prompt": true
    }
   ]
  },
  {
   "voice": "Matthew Berman",
   "records": 15,
   "models": 6,
   "early_access_share": 0.53,
   "domains": {
    "general": 10,
    "agentic": 2,
    "frontend-design": 1,
    "coding": 1,
    "multimodal": 1
   },
   "scores": {
    "Claude Fable 5.2": 1.9,
    "DeepSeek-V4.1-Flash-1017": 1.75,
    "Kimi K3.1 Pro": 0.17,
    "Gemini 4 Carbon": 0.0,
    "GPT-6.2 Astra": 0.0,
    "GPT-6.2 Sol": 0.0
   },
   "rank_correlation": [
    {
     "benchmark": "DeckBench",
     "rho": 0.39,
     "n": 6,
     "spring_prompt": true
    },
    {
     "benchmark": "DesignArena Overall Agentic",
     "rho": 0.34,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "SurveyBench",
     "rho": 0.21,
     "n": 6,
     "spring_prompt": true
    },
    {
     "benchmark": "ROASBench",
     "rho": 0.21,
     "n": 6,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (factual brief)",
     "rho": 0.15,
     "n": 6,
     "spring_prompt": true
    },
    {
     "benchmark": "Overall leaderboard (Spring Prompt, blends third-party sources)",
     "rho": 0.09,
     "n": 6,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 long_agentic",
     "rho": 0.03,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 design",
     "rho": 0.03,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "Vals Index",
     "rho": -0.03,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "Terminal-Bench 4.1 (tbench.ai)",
     "rho": -0.03,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "Spring Prompt Score",
     "rho": -0.03,
     "n": 6,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 research",
     "rho": -0.03,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena WebDev",
     "rho": -0.03,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena Agent (Agent Arena Index)",
     "rho": -0.03,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "AA Intelligence Index",
     "rho": -0.03,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 overall",
     "rho": -0.09,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "CatalogBench (sales brief) [reliable]",
     "rho": -0.45,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (factual brief) [reliable]",
     "rho": -0.45,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 data_analysis",
     "rho": -0.46,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "CatalogBench (sales brief)",
     "rho": -0.46,
     "n": 6,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 browser",
     "rho": -0.52,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 reasoning",
     "rho": -0.71,
     "n": 6,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 coding",
     "rho": -0.76,
     "n": 6,
     "spring_prompt": false
    }
   ]
  },
  {
   "voice": "ThePrimeagen",
   "records": 8,
   "models": 6,
   "early_access_share": 0.0,
   "domains": {
    "agentic": 3,
    "general": 2,
    "coding": 2,
    "multimodal": 1
   },
   "scores": {
    "GPT-6.2 Sol": 1.5,
    "Muse Ember": 1.5,
    "Gemini 4 Argon": 1.5,
    "GPT-6.2 Astra": 1.0,
    "DeepSeek-V4.1-Flash-1017": 0.5,
    "Claude Fable 5.2": -0.5
   },
   "rank_correlation": [
    {
     "benchmark": "CatalogBench (sales brief) [reliable]",
     "rho": 0.36,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (sales brief)",
     "rho": 0.36,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (factual brief) [reliable]",
     "rho": 0.36,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "DeckBench",
     "rho": -0.21,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (factual brief)",
     "rho": -0.21,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "SurveyBench",
     "rho": -0.36,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "Spring Prompt Score",
     "rho": -0.36,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "ROASBench",
     "rho": -0.36,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "Overall leaderboard (Spring Prompt, blends third-party sources)",
     "rho": -0.36,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "AA Intelligence Index",
     "rho": -0.36,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena Agent (Agent Arena Index)",
     "rho": -0.5,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "Vals Index",
     "rho": -0.62,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "Terminal-Bench 4.1 (tbench.ai)",
     "rho": -0.62,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena WebDev",
     "rho": -0.62,
     "n": 5,
     "spring_prompt": false
    }
   ]
  },
  {
   "voice": "Every",
   "records": 29,
   "models": 5,
   "early_access_share": 0.9,
   "domains": {
    "general": 10,
    "coding": 8,
    "writing": 5,
    "frontend-design": 3,
    "agentic": 2,
    "price-speed": 1
   },
   "scores": {
    "Gemini 4 Argon": 1.5,
    "Claude Fable 5.2": 1.38,
    "Gemini 4 Carbon": 0.36,
    "GPT-6.2 Sol": -0.5,
    "GPT-6.2 Astra": -0.95
   },
   "rank_correlation": []
  },
  {
   "voice": "Ethan Mollick",
   "records": 10,
   "models": 5,
   "early_access_share": 0.9,
   "domains": {
    "general": 6,
    "research": 2,
    "agentic": 2
   },
   "scores": {
    "Gemini 4 Carbon": 1.0,
    "Claude Fable 5.2": 0.88,
    "Kimi K3.1 Pro": 0.0,
    "GPT-6.2 Astra": -0.67,
    "GPT-6.2 Sol": -1.0
   },
   "rank_correlation": [
    {
     "benchmark": "Vals Index",
     "rho": 0.6,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "Terminal-Bench 4.1 (tbench.ai)",
     "rho": 0.6,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 research",
     "rho": 0.6,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 long_agentic",
     "rho": 0.6,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena Agent (Agent Arena Index)",
     "rho": 0.6,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 data_analysis",
     "rho": 0.5,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 design",
     "rho": 0.3,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "AA Intelligence Index",
     "rho": 0.3,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 overall",
     "rho": 0.21,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "SurveyBench",
     "rho": -0.1,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "Spring Prompt Score",
     "rho": -0.1,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "ROASBench",
     "rho": -0.1,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 coding",
     "rho": -0.1,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "Arena WebDev",
     "rho": -0.1,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "Overall leaderboard (Spring Prompt, blends third-party sources)",
     "rho": -0.3,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "DeckBench",
     "rho": -0.3,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "CatalogBench (factual brief)",
     "rho": -0.36,
     "n": 5,
     "spring_prompt": true
    },
    {
     "benchmark": "LiveBench 2.0 reasoning",
     "rho": -0.41,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "LiveBench 2.0 browser",
     "rho": -0.6,
     "n": 5,
     "spring_prompt": false
    },
    {
     "benchmark": "CatalogBench (sales brief)",
     "rho": -0.97,
     "n": 5,
     "spring_prompt": true
    }
   ]
  },
  {
   "voice": "Prompt Engineering",
   "records": 8,
   "models": 4,
   "early_access_share": 0.25,
   "domains": {
    "agentic": 4,
    "price-speed": 2,
    "multimodal": 1,
    "frontend-design": 1
   },
   "scores": {
    "Gemini 4 Neon": 1.5,
    "Gemini 4 Carbon": 1.0,
    "Gemini Decision 1": 0.75,
    "Kimi K3.1 Pro": 0.17
   },
   "rank_correlation": []
  },
  {
   "voice": "Paweł Huryn",
   "records": 4,
   "models": 4,
   "early_access_share": 0.0,
   "domains": {
    "coding": 4
   },
   "scores": {
    "GPT-6.2 Astra": 2.0,
    "Gemini 4 Carbon": 1.0,
    "Gemini 4 Hydrogen": 1.0,
    "GPT-6.1 Terra": 0.0
   },
   "rank_correlation": []
  },
  {
   "voice": "Chetaslua",
   "records": 5,
   "models": 4,
   "early_access_share": 0.0,
   "domains": {
    "frontend-design": 3,
    "general": 1,
    "multimodal": 1
   },
   "scores": {
    "Gemini 4 Carbon": 2.0,
    "GPT-6.2 Astra": 2.0,
    "Gemini 4 Neon": 1.0,
    "Claude Fable 5.2": 1.0
   },
   "rank_correlation": []
  },
  {
   "voice": "Pankaj Kumar",
   "records": 5,
   "models": 4,
   "early_access_share": 0.0,
   "domains": {
    "frontend-design": 3,
    "multimodal": 2
   },
   "scores": {
    "Gemini 4 Hydrogen": 2.0,
    "Claude Fable 5.2": 2.0,
    "GPT-6.2 Astra": 1.0,
    "Gemini 4 Carbon": 1.0
   },
   "rank_correlation": []
  },
  {
   "voice": "Peter Yang",
   "records": 18,
   "models": 3,
   "early_access_share": 0.33,
   "domains": {
    "frontend-design": 7,
    "writing": 6,
    "general": 3,
    "research": 1,
    "coding": 1
   },
   "scores": {
    "Claude Fable 5.2": 0.88,
    "Gemini 4 Carbon": 0.38,
    "GPT-6.2 Astra": -0.33
   },
   "rank_correlation": []
  },
  {
   "voice": "Wes Roth",
   "records": 6,
   "models": 3,
   "early_access_share": 0.0,
   "domains": {
    "frontend-design": 2,
    "coding": 1,
    "multimodal": 1,
    "agentic": 1,
    "writing": 1
   },
   "scores": {
    "Gemini 4 Carbon": 2.0,
    "Claude Fable 5.2": 1.75,
    "GPT-6.2 Astra": 0.0
   },
   "rank_correlation": []
  },
  {
   "voice": "Matt Shumer",
   "records": 11,
   "models": 3,
   "early_access_share": 0.91,
   "domains": {
    "general": 6,
    "agentic": 3,
    "coding": 1,
    "multimodal": 1
   },
   "scores": {
    "Claude Fable 5.2": 0.6,
    "GPT-6.2 Astra": -2.0,
    "GPT-6.2 Sol": -2.0
   },
   "rank_correlation": []
  },
  {
   "voice": "Flavio Adamo",
   "records": 7,
   "models": 3,
   "early_access_share": 0.57,
   "domains": {
    "multimodal": 2,
    "frontend-design": 2,
    "general": 1,
    "coding": 1,
    "writing": 1
   },
   "scores": {
    "Claude Fable 5.2": 1.83,
    "GPT-6.1 Terra": 1.0,
    "GPT-6.2 Astra": 0.5
   },
   "rank_correlation": []
  },
  {
   "voice": "Pietro Schirano",
   "records": 4,
   "models": 3,
   "early_access_share": 0.5,
   "domains": {
    "general": 2,
    "frontend-design": 1,
    "price-speed": 1
   },
   "scores": {
    "GPT-6.2 Astra": 1.25,
    "GPT-6.1 Terra": 1.0,
    "Claude Fable 5.2": 0.0
   },
   "rank_correlation": []
  },
  {
   "voice": "Sam Witteveen",
   "records": 4,
   "models": 2,
   "early_access_share": 0.5,
   "domains": {
    "coding": 1,
    "frontend-design": 1,
    "price-speed": 1,
    "writing": 1
   },
   "scores": {
    "Gemini 4 Carbon": 1.5,
    "Mercury 3": 1.25
   },
   "rank_correlation": []
  },
  {
   "voice": "Fireship",
   "records": 2,
   "models": 2,
   "early_access_share": 0.0,
   "domains": {
    "frontend-design": 2
   },
   "scores": {
    "Claude Fable 5.2": 1.5,
    "Kimi K3.1 Pro": -0.5
   },
   "rank_correlation": []
  },
  {
   "voice": "WorldofAI",
   "records": 3,
   "models": 2,
   "early_access_share": 0.67,
   "domains": {
    "frontend-design": 2,
    "general": 1
   },
   "scores": {
    "Gemini 4 Neon": 1.0,
    "Qwen4-Max": 0.25
   },
   "rank_correlation": []
  },
  {
   "voice": "McKay Wrigley",
   "records": 3,
   "models": 2,
   "early_access_share": 0.0,
   "domains": {
    "general": 3
   },
   "scores": {
    "GPT-6.2 Sol": 1.0,
    "GPT-6.2 Astra": -1.75
   },
   "rank_correlation": []
  },
  {
   "voice": "AI Explained",
   "records": 1,
   "models": 1,
   "early_access_share": 1.0,
   "domains": {
    "agentic": 1
   },
   "scores": {
    "Claude Fable 5.2": -1.0
   },
   "rank_correlation": []
  },
  {
   "voice": "Zvi Mowshowitz",
   "records": 5,
   "models": 1,
   "early_access_share": 0.0,
   "domains": {
    "writing": 2,
    "general": 1,
    "agentic": 1,
    "coding": 1
   },
   "scores": {
    "Claude Fable 5.2": -0.1
   },
   "rank_correlation": []
  },
  {
   "voice": "Claire Vo",
   "records": 3,
   "models": 1,
   "early_access_share": 0.0,
   "domains": {
    "general": 1,
    "coding": 1,
    "frontend-design": 1
   },
   "scores": {
    "Claude Fable 5.2": 1.0
   },
   "rank_correlation": []
  },
  {
   "voice": "The Neuron",
   "records": 1,
   "models": 1,
   "early_access_share": 0.0,
   "domains": {
    "general": 1
   },
   "scores": {
    "Mercury 3": 2.0
   },
   "rank_correlation": []
  },
  {
   "voice": "Riley Brown",
   "records": 1,
   "models": 1,
   "early_access_share": 0.0,
   "domains": {
    "multimodal": 1
   },
   "scores": {
    "Claude Fable 5.2": 2.0
   },
   "rank_correlation": []
  },
  {
   "voice": "Deedy Das",
   "records": 1,
   "models": 1,
   "early_access_share": 0.0,
   "domains": {
    "frontend-design": 1
   },
   "scores": {
    "Claude Fable 5.2": 2.0
   },
   "rank_correlation": []
  }
 ]
}