{
  "type": "tweet",
  "id": "2072066402710618322",
  "url": "https://x.com/IBuzovskyi/status/2072066402710618322",
  "twitterUrl": "https://twitter.com/IBuzovskyi/status/2072066402710618322",
  "text": "HERMES AGENT NOW READS THE WEB\nUP TO 60X FASTER AND 49X CHEAPER.\nCLEAN CONTENT STRAIGHT TO THE AGENT.\nLARGE PAGES PAGED ON DEMAND.\n\n@NousResearch \n\nscraping backends used to return raw content\nthat got processed redundantly before\nreaching the agent. that pipeline is gone.\n\nnow: backends pass clean content directly.\nlarge pages save locally and page on demand.\nsame quality. fraction of the time and cost.\n\nHOW WEB_EXTRACT HANDLES LARGE PAGES:\n\nsize-driven processing. no wasted tokens.\n\nunder 5,000 chars:\n→ returned as-is. no LLM call.\n  full markdown reaches the agent.\n\n5,000 to 500,000 chars:\n→ single-pass summary via auxiliary model.\n  capped at ~5,000 chars of output.\n  keeps quotes, code blocks, key facts.\n\n500,000 to 2,000,000 chars:\n→ chunked into 100K-char pieces.\n  each chunk summarized in parallel.\n  final synthesis: ~5,000 chars.\n\nover 2,000,000 chars:\n→ refused with a hint to use web_crawl\n  with focused extraction instructions.\n\nthe summary is a content compressor, not a paraphraser.\nif summarization fails, Hermes falls back\nto the first ~5,000 chars of raw content.\nno useless error messages.\n\nROUTE EXTRACTION TO A CHEAP MODEL:\n\nby default, web_extract uses your main model.\non Opus that means every long page\nburns premium tokens on summarization.\n\nset in Desktop app, Dashboard, or config.yaml:\n\nauxiliary:\n  web_extract:\n    provider: openrouter\n    model: google/gemini-3-flash-preview\n    timeout: 360\n\nextraction summaries on Gemini Flash.\nreasoning stays on your premium model.\nthis alone cuts web research costs significantly.\n\n8 BACKEND PROVIDERS:\n\nFirecrawl (default): search + extract + crawl. 500 free credits/month.\nSearXNG: free, self-hosted, search-only. no API key.\nBrave Search: 2,000 free queries/month. search-only.\nDDGS (DuckDuckGo): free, no key needed. search-only.\nTavily: search + extract + crawl. 1,000 free searches/month.\nExa: search + extract. 1,000 free searches/month.\nParallel: search + extract. paid.\nxAI (Grok): search-only. LLM-generated results via Grok.\n\nsearch-only providers pair with Firecrawl/Tavily/Exa\nfor extract capability.\n\nPER-CAPABILITY SPLIT:\n\nuse different providers for search vs extract:\n\nSearXNG (free) for search.\nFirecrawl for extract.\n\nfree searches. paid extraction only when needed.\nconfigure via hermes tools or config.yaml.\n\nFREE SELF-HOSTED SEARCH (SEARXNG):\n\nzero API costs. zero rate limits.\nprivacy-respecting metasearch across 70+ engines.\n\ndocker compose up -d\n\nset SEARXNG_URL in .env.\nenable JSON format in settings.yml.\nHermes connects automatically.\n\npair with Firecrawl for extract\nand you have search for free\nwith paid extraction only on demand.\n\nNOUS PORTAL SUBSCRIBERS:\n\nweb search and extract included\nthrough the Tool Gateway via managed Firecrawl.\nno API key needed. no separate billing.\nhermes setup --portal enables everything.\n\nWHEN YOU NEED RAW CONTENT:\n\nif the LLM summary drops important fields\n(structured data, tables, specific formatting):\n\nuse browser_navigate + browser_snapshot instead.\nreturns the live accessibility tree\nwithout auxiliary-model rewriting.\n\nfull Hermes architecture deep-dive in the article 👇",
  "source": "Twitter for iPhone",
  "retweetCount": 2,
  "replyCount": 2,
  "likeCount": 47,
  "quoteCount": 0,
  "viewCount": 4331,
  "createdAt": "Tue Jun 30 21:15:02 +0000 2026",
  "lang": "en",
  "bookmarkCount": 42,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2072066402710618322",
  "displayTextRange": [
    0,
    279
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "IBuzovskyi",
    "url": "https://x.com/IBuzovskyi",
    "twitterUrl": "https://twitter.com/IBuzovskyi",
    "id": "1387382307229769732",
    "name": "YanXbt",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/2055755733002506240/uFECjFhT_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/1387382307229769732/1779712969",
    "description": "",
    "location": "San Francisco, CA",
    "followers": 2813,
    "following": 1088,
    "status": "",
    "canDm": true,
    "canMediaTag": true,
    "createdAt": "Wed Apr 28 12:25:00 +0000 2021",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 3528,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 708,
    "statusesCount": 2905,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "2071601107944571249"
    ],
    "profile_bio": {
      "description": "Content creator | AI Systems builder | Zero-headcount believer\nTeaching how to build with HERMES AGENT",
      "entities": {
        "description": {},
        "url": {
          "urls": [
            {
              "display_url": "substack.com/@yanxbt",
              "expanded_url": "https://substack.com/@yanxbt",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/ZyqLGGM9Ku"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "additional_media_info": {
          "monetizable": false
        },
        "allow_download_status": {
          "allow_download": true
        },
        "display_url": "pic.twitter.com/MnmlVyUwmQ",
        "expanded_url": "https://twitter.com/IBuzovskyi/status/2072066402710618322/video/1",
        "ext_master_playlist_only": [],
        "ext_media_availability": {
          "status": "Available"
        },
        "ext_playlists": [],
        "id_str": "2072065936043986945",
        "indices": [
          280,
          303
        ],
        "media_key": "13_2072065936043986945",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwABAoAARzBdP2VFtABAAA=",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAAECgABHMF0/ZUW0AEAAA==",
            "media_key": "13_2072065936043986945"
          }
        },
        "media_url_https": "https://pbs.twimg.com/amplify_video_thumb/2072065936043986945/img/XXs5bm4FvGFheM4K.jpg",
        "original_info": {
          "focus_rects": [],
          "height": 2160,
          "width": 2160
        },
        "sizes": {
          "large": {
            "h": 2048,
            "w": 2048
          }
        },
        "type": "video",
        "url": "https://t.co/MnmlVyUwmQ",
        "video_info": {
          "aspect_ratio": [
            1,
            1
          ],
          "duration_millis": 13003,
          "variants": [
            {
              "content_type": "application/x-mpegURL",
              "url": "https://video.twimg.com/amplify_video/2072065936043986945/pl/kjaVMJFHhqG8oXct.m3u8?tag=28&v=cfc"
            },
            {
              "bitrate": 432000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2072065936043986945/vid/avc1/320x320/fDb_Whf4q2MtoXv7.mp4?tag=28"
            },
            {
              "bitrate": 832000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2072065936043986945/vid/avc1/540x540/HgoBAu_Yb61IAZR_.mp4?tag=28"
            },
            {
              "bitrate": 1280000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2072065936043986945/vid/avc1/720x720/7ES5TPn3XZ5ETgWI.mp4?tag=28"
            },
            {
              "bitrate": 8768000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2072065936043986945/vid/avc1/1080x1080/lBIx4BeZBb1aIEpQ.mp4?tag=28"
            },
            {
              "bitrate": 25128000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2072065936043986945/vid/avc1/2160x2160/BzibPkb8r2VNlqDX.mp4?tag=28"
            }
          ]
        }
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "symbols": [],
    "timestamps": [],
    "urls": [],
    "user_mentions": [
      {
        "id_str": "1318419526132862976",
        "indices": [
          132,
          145
        ],
        "name": "Nous Research",
        "screen_name": "NousResearch"
      }
    ]
  },
  "quoted_tweet": {
    "type": "tweet",
    "id": "2068629714776756339",
    "url": "https://x.com/IBuzovskyi/status/2068629714776756339",
    "twitterUrl": "https://twitter.com/IBuzovskyi/status/2068629714776756339",
    "text": "https://t.co/VxyyeQCimO",
    "source": "Twitter for iPhone",
    "retweetCount": 61,
    "replyCount": 14,
    "likeCount": 487,
    "quoteCount": 5,
    "viewCount": 365140,
    "createdAt": "Sun Jun 21 09:38:51 +0000 2026",
    "lang": "zxx",
    "bookmarkCount": 1228,
    "isReply": false,
    "inReplyToId": null,
    "conversationId": "2068629714776756339",
    "displayTextRange": [
      0,
      23
    ],
    "inReplyToUserId": null,
    "inReplyToUsername": null,
    "author": {
      "type": "user",
      "userName": "IBuzovskyi",
      "url": "https://x.com/IBuzovskyi",
      "twitterUrl": "https://twitter.com/IBuzovskyi",
      "id": "1387382307229769732",
      "name": "YanXbt",
      "isVerified": false,
      "isBlueVerified": true,
      "verifiedType": null,
      "profilePicture": "https://pbs.twimg.com/profile_images/2055755733002506240/uFECjFhT_normal.jpg",
      "coverPicture": "https://pbs.twimg.com/profile_banners/1387382307229769732/1779712969",
      "description": "",
      "location": "San Francisco, CA",
      "followers": 2813,
      "following": 1088,
      "status": "",
      "canDm": true,
      "canMediaTag": true,
      "createdAt": "Wed Apr 28 12:25:00 +0000 2021",
      "entities": {
        "description": {
          "urls": []
        },
        "url": {}
      },
      "fastFollowersCount": 0,
      "favouritesCount": 3528,
      "hasCustomTimelines": true,
      "isTranslator": false,
      "mediaCount": 708,
      "statusesCount": 2905,
      "withheldInCountries": [],
      "affiliatesHighlightedLabel": {},
      "possiblySensitive": false,
      "pinnedTweetIds": [
        "2071601107944571249"
      ],
      "profile_bio": {
        "description": "Content creator | AI Systems builder | Zero-headcount believer\nTeaching how to build with HERMES AGENT",
        "entities": {
          "description": {},
          "url": {
            "urls": [
              {
                "display_url": "substack.com/@yanxbt",
                "expanded_url": "https://substack.com/@yanxbt",
                "indices": [
                  0,
                  23
                ],
                "url": "https://t.co/ZyqLGGM9Ku"
              }
            ]
          }
        }
      },
      "isAutomated": false,
      "automatedBy": null
    },
    "extendedEntities": {},
    "card": null,
    "place": {},
    "entities": {
      "urls": [
        {
          "display_url": "x.com/i/article/2068…",
          "expanded_url": "http://x.com/i/article/2068585191522566144",
          "indices": [
            0,
            23
          ],
          "url": "https://t.co/VxyyeQCimO"
        }
      ]
    },
    "quoted_tweet": null,
    "retweeted_tweet": null,
    "isLimitedReply": false,
    "communityInfo": null,
    "article": {
      "title": "15 LEVELS OF HERMES AGENT. FROM CHATBOT TO 24/7 AUTONOMOUS SYSTEM.",
      "preview_text": "Most people install Hermes Agent and use it as a chatbot. They type a prompt, get a response, close the tab. That covers maybe 10% of what the agent can do.\nThis article maps every level of Hermes",
      "cover_media_img_url": "https://pbs.twimg.com/media/HLU_q3TXUAAmoT2.jpg"
    }
  },
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "communityInfo": null,
  "article": null
}