🐦 Twitter Post Details

Viewing enriched Twitter post

@bojie_li

Closed labs hide model sizes. They can't hide what their models know, and what a model knows is an indicator on how big it is. Reasoning compresses. Factual knowledge doesn't. So you can size a frontier model from black-box API calls alone, and across releases you can literally watch a single fact arrive in the parameters over time. For three years, my friends Jiyan He and Zihan Zheng have been asking frontier LLMs the same question: "what do you know about USTC Hackergame?", a CTF contest. May 2024: GPT-4o invented fake titles. Feb 2025: Claude 3.7 Sonnet listed 19 verified 2023 challenges. By April 2026, frontier models recall specific challenges across consecutive years. After DeepSeek-V4 dropped, I instructed my agent to spend four days autonomously turning that habit into Incompressible Knowledge Probes (IKP) — 1,400 questions, 7 tiers of obscurity, 188 models, 27 vendors. Three findings: 1/ You can approximately size any black-box LLM from factual accuracy alone. Penalized accuracy is log-linear in log(params), R² = 0.917 on 89 open-weight models from 135M to 1.6T params. Project closed APIs onto the curve → GPT-5.5 ~9T, Claude Opus 4.7 ~4T, GPT-5.4 ~2.2T, Claude Sonnet 4.6 ~1.7T, Gemini 2.5 Pro ~1.2T (90% CI: 0.3-3x size). 2/ Citation count and h-index don't predict whether a frontier model recognizes a researcher. Two researchers with similar citation profiles get very different responses. Models memorize impact — work that shaped a field, not many incremental papers. 3/ Factual capacity doesn't compress over time. Across 96 open-weight models across 3 years, the IKP time coefficient is statistically zero, rejecting the Densing-Law prediction of +0.0117/month at p<10⁻¹⁵. Reasoning benchmarks saturate; factual capacity keeps scaling with parameters. Website: https://t.co/CkwJsXqnsX Paper: https://t.co/eNUdC9ye7w

Media 1
Media 2
Media 3

📊 Media Metadata

{
  "media": [
    {
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2049314403208896521/media_0.jpg",
      "media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2049314403208896521/media_0.jpg",
      "type": "photo",
      "filename": "media_0.jpg"
    },
    {
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2049314403208896521/media_1.jpg",
      "media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2049314403208896521/media_1.jpg",
      "type": "photo",
      "filename": "media_1.jpg"
    },
    {
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2049314403208896521/media_2.jpg",
      "media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2049314403208896521/media_2.jpg",
      "type": "photo",
      "filename": "media_2.jpg"
    }
  ],
  "processed_at": "2026-05-11T22:39:09.779098",
  "pipeline_version": "2.0"
}

🔧 Raw API Response

{
  "type": "tweet",
  "id": "2049314403208896521",
  "url": "https://x.com/bojie_li/status/2049314403208896521",
  "twitterUrl": "https://twitter.com/bojie_li/status/2049314403208896521",
  "text": "Closed labs hide model sizes. They can't hide what their models know, and what a model knows is an indicator on how big it is.\n\nReasoning compresses. Factual knowledge doesn't. So you can size a frontier model from black-box API calls alone, and across releases you can literally watch a single fact arrive in the parameters over time.\n\nFor three years, my friends Jiyan He and Zihan Zheng have been asking frontier LLMs the same question: \"what do you know about USTC Hackergame?\", a CTF contest. May 2024: GPT-4o invented fake titles. Feb 2025: Claude 3.7 Sonnet listed 19 verified 2023 challenges. By April 2026, frontier models recall specific challenges across consecutive years.\n\nAfter DeepSeek-V4 dropped, I instructed my agent to spend four days autonomously turning that habit into Incompressible Knowledge Probes (IKP) — 1,400 questions, 7 tiers of obscurity, 188 models, 27 vendors. Three findings:\n\n1/ You can approximately size any black-box LLM from factual accuracy alone. Penalized accuracy is log-linear in log(params), R² = 0.917 on 89 open-weight models from 135M to 1.6T params. Project closed APIs onto the curve → GPT-5.5 ~9T, Claude Opus 4.7 ~4T, GPT-5.4 ~2.2T, Claude Sonnet 4.6 ~1.7T, Gemini 2.5 Pro ~1.2T (90% CI: 0.3-3x size).\n\n2/ Citation count and h-index don't predict whether a frontier model recognizes a researcher. Two researchers with similar citation profiles get very different responses. Models memorize impact — work that shaped a field, not many incremental papers.\n\n3/ Factual capacity doesn't compress over time. Across 96 open-weight models across 3 years, the IKP time coefficient is statistically zero, rejecting the Densing-Law prediction of +0.0117/month at p<10⁻¹⁵. Reasoning benchmarks saturate; factual capacity keeps scaling with parameters.\n\nWebsite: https://t.co/CkwJsXqnsX\nPaper: https://t.co/eNUdC9ye7w",
  "source": "Twitter for iPhone",
  "retweetCount": 234,
  "replyCount": 71,
  "likeCount": 2189,
  "quoteCount": 54,
  "viewCount": 386682,
  "createdAt": "Wed Apr 29 02:26:42 +0000 2026",
  "lang": "en",
  "bookmarkCount": 1390,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2049314403208896521",
  "displayTextRange": [
    0,
    279
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "bojie_li",
    "url": "https://x.com/bojie_li",
    "twitterUrl": "https://twitter.com/bojie_li",
    "id": "1479100058",
    "name": "Bojie Li",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/819210652564615168/NNNeTcef_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/1479100058/1696192511",
    "description": "",
    "location": "",
    "followers": 1274,
    "following": 559,
    "status": "",
    "canDm": false,
    "canMediaTag": true,
    "createdAt": "Mon Jun 03 07:31:22 +0000 2013",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 563,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 35,
    "statusesCount": 170,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [],
    "profile_bio": {
      "description": "Chief Scientist, Pine AI",
      "entities": {
        "description": {},
        "url": {
          "urls": [
            {
              "display_url": "01.me",
              "expanded_url": "https://01.me/",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/9bsINBiPaj"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "display_url": "pic.twitter.com/Yab75gkWNW",
        "expanded_url": "https://twitter.com/bojie_li/status/2049314403208896521/photo/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "features": {
          "large": {
            "faces": []
          },
          "orig": {
            "faces": []
          }
        },
        "id_str": "2049307408753975296",
        "indices": [
          280,
          303
        ],
        "media_key": "3_2049307408753975296",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARxwmjsGW2AACgACHHCgl4xboAkAAA==",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAABCgABHHCaOwZbYAAKAAIccKCXjFugCQAA",
            "media_key": "3_2049307408753975296"
          }
        },
        "media_url_https": "https://pbs.twimg.com/media/HHCaOwZbYAAmGyk.jpg",
        "original_info": {
          "focus_rects": [
            {
              "h": 1245,
              "w": 2224,
              "x": 0,
              "y": 157
            },
            {
              "h": 1402,
              "w": 1402,
              "x": 0,
              "y": 0
            },
            {
              "h": 1402,
              "w": 1230,
              "x": 0,
              "y": 0
            },
            {
              "h": 1402,
              "w": 701,
              "x": 0,
              "y": 0
            },
            {
              "h": 1402,
              "w": 2224,
              "x": 0,
              "y": 0
            }
          ],
          "height": 1402,
          "width": 2224
        },
        "sizes": {
          "large": {
            "h": 1291,
            "w": 2048
          }
        },
        "type": "photo",
        "url": "https://t.co/Yab75gkWNW"
      },
      {
        "display_url": "pic.twitter.com/Yab75gkWNW",
        "expanded_url": "https://twitter.com/bojie_li/status/2049314403208896521/photo/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "features": {
          "large": {
            "faces": []
          },
          "orig": {
            "faces": []
          }
        },
        "id_str": "2049307408762327040",
        "indices": [
          280,
          303
        ],
        "media_key": "3_2049307408762327040",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARxwmjsG2tAACgACHHCgl4xboAkAAA==",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAABCgABHHCaOwba0AAKAAIccKCXjFugCQAA",
            "media_key": "3_2049307408762327040"
          }
        },
        "media_url_https": "https://pbs.twimg.com/media/HHCaOwba0AA6kuA.jpg",
        "original_info": {
          "focus_rects": [
            {
              "h": 997,
              "w": 1780,
              "x": 0,
              "y": 0
            },
            {
              "h": 1328,
              "w": 1328,
              "x": 447,
              "y": 0
            },
            {
              "h": 1328,
              "w": 1165,
              "x": 529,
              "y": 0
            },
            {
              "h": 1328,
              "w": 664,
              "x": 779,
              "y": 0
            },
            {
              "h": 1328,
              "w": 1780,
              "x": 0,
              "y": 0
            }
          ],
          "height": 1328,
          "width": 1780
        },
        "sizes": {
          "large": {
            "h": 1328,
            "w": 1780
          }
        },
        "type": "photo",
        "url": "https://t.co/Yab75gkWNW"
      },
      {
        "display_url": "pic.twitter.com/Yab75gkWNW",
        "expanded_url": "https://twitter.com/bojie_li/status/2049314403208896521/photo/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "features": {
          "large": {
            "faces": []
          },
          "orig": {
            "faces": []
          }
        },
        "id_str": "2049307991548239872",
        "indices": [
          280,
          303
        ],
        "media_key": "3_2049307991548239872",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARxwmsK3miAACgACHHCgl4xboAkAAA==",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAABCgABHHCawreaIAAKAAIccKCXjFugCQAA",
            "media_key": "3_2049307991548239872"
          }
        },
        "media_url_https": "https://pbs.twimg.com/media/HHCawreaIAAEcph.jpg",
        "original_info": {
          "focus_rects": [
            {
              "h": 640,
              "w": 1142,
              "x": 0,
              "y": 0
            },
            {
              "h": 954,
              "w": 954,
              "x": 0,
              "y": 0
            },
            {
              "h": 954,
              "w": 837,
              "x": 10,
              "y": 0
            },
            {
              "h": 954,
              "w": 477,
              "x": 190,
              "y": 0
            },
            {
              "h": 954,
              "w": 1142,
              "x": 0,
              "y": 0
            }
          ],
          "height": 954,
          "width": 1142
        },
        "sizes": {
          "large": {
            "h": 954,
            "w": 1142
          }
        },
        "type": "photo",
        "url": "https://t.co/Yab75gkWNW"
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "symbols": [],
    "urls": [
      {
        "display_url": "01.me/research/ikp/",
        "expanded_url": "https://01.me/research/ikp/",
        "indices": [
          1803,
          1826
        ],
        "url": "https://t.co/CkwJsXqnsX"
      },
      {
        "display_url": "arxiv.org/pdf/2604.24827",
        "expanded_url": "https://arxiv.org/pdf/2604.24827",
        "indices": [
          1834,
          1857
        ],
        "url": "https://t.co/eNUdC9ye7w"
      }
    ],
    "user_mentions": []
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "communityInfo": null,
  "article": null
}