{
  "type": "tweet",
  "id": "1980722682246398069",
  "url": "https://x.com/arankomatsuzaki/status/1980722682246398069",
  "twitterUrl": "https://twitter.com/arankomatsuzaki/status/1980722682246398069",
  "text": "DeepSeek OCR dropped ... but honestly, Glyph [1], released the same day, showed something more interesting: 3–4× context compression and infilling cost reduction, no performance hit on long-context QA and summarization, which is much less trivial than OCR in many cases.\n\nIf that holds for harder agentic tasks, that’s a serious leap.\n\nCost-wise:\n\n- Infilling cost drops sharply\n- Decoding savings are more modest w/ DSA on\n\nSo the impact depends on how input-heavy your agentic workflow is (e.g., deep research vs coding from scratch).\n\nAlso relevant:\n\n- BLT extensions [2,3] improved scaling over BPE baseline; aggressive compression of Glyph mainly helps infilling, not much on decoding (w/ DSA).\n- BLT-fication could help Glyph reduce decoding cost further.\n- Subagents make bigger impact on latency and context length reduction. Simple yet powerful.\n- And swapping vision encoders for small LMs is still an open question.",
  "source": "Twitter for iPhone",
  "retweetCount": 35,
  "replyCount": 4,
  "likeCount": 220,
  "quoteCount": 0,
  "viewCount": 24975,
  "createdAt": "Tue Oct 21 19:47:42 +0000 2025",
  "lang": "en",
  "bookmarkCount": 136,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "1980722682246398069",
  "displayTextRange": [
    0,
    280
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "arankomatsuzaki",
    "url": "https://x.com/arankomatsuzaki",
    "twitterUrl": "https://twitter.com/arankomatsuzaki",
    "id": "794433401591693312",
    "name": "Aran Komatsuzaki",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/1561220982328754176/JOYS5kab_normal.jpg",
    "coverPicture": "",
    "description": "",
    "location": "",
    "followers": 151550,
    "following": 315,
    "status": "",
    "canDm": true,
    "canMediaTag": true,
    "createdAt": "Fri Nov 04 06:57:37 +0000 2016",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 15010,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 2525,
    "statusesCount": 6496,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [],
    "profile_bio": {
      "description": "Sharing AI research. Early work on AI (GPT-J, LAION, scaling, MoE). Ex ML PhD (GT) & Google.",
      "entities": {
        "description": {},
        "url": {
          "urls": [
            {
              "display_url": "arankomatsuzaki.wordpress.com/about-me/",
              "expanded_url": "https://arankomatsuzaki.wordpress.com/about-me/",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/aZGCShojNY"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "allow_download_status": {
          "allow_download": true
        },
        "display_url": "pic.twitter.com/BYbfktijzO",
        "expanded_url": "https://twitter.com/arankomatsuzaki/status/1980722682246398069/photo/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "features": {
          "large": {},
          "orig": {}
        },
        "id_str": "1980722445125332992",
        "indices": [
          281,
          304
        ],
        "media_key": "3_1980722445125332992",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARt88JHMVrAACgACG3zwyQHbAHUAAA==",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAABCgABG3zwkcxWsAAKAAIbfPDJAdsAdQAA",
            "media_key": "3_1980722445125332992"
          }
        },
        "media_url_https": "https://pbs.twimg.com/media/G3zwkcxWsAAcM4e.jpg",
        "original_info": {
          "focus_rects": [
            {
              "h": 576,
              "w": 1029,
              "x": 0,
              "y": 0
            },
            {
              "h": 576,
              "w": 576,
              "x": 109,
              "y": 0
            },
            {
              "h": 576,
              "w": 505,
              "x": 145,
              "y": 0
            },
            {
              "h": 576,
              "w": 288,
              "x": 253,
              "y": 0
            },
            {
              "h": 576,
              "w": 1060,
              "x": 0,
              "y": 0
            }
          ],
          "height": 576,
          "width": 1060
        },
        "sizes": {
          "large": {
            "h": 576,
            "w": 1060
          }
        },
        "type": "photo",
        "url": "https://t.co/BYbfktijzO"
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {},
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "article": null
}