🐦 Twitter Post Details

Viewing enriched Twitter post

@dair_ai

New research from Georgia Tech and Microsoft Research. GUI agents today are reactive. Every step costs an LLM call, which is why a lot of GUI agents are expensive, slow, and fragile. This new research introduces ActionEngine, a framework that shifts GUI agents from reactive execution to programmatic planning. A Crawling Agent explores the application offline and builds a state-machine graph of the interface. Nodes are page states, edges are actions. Then at runtime, an Execution Agent uses this graph to synthesize a complete Python program in a single LLM call. Instead of O(N) vision model calls per task, you get O(1) planning cost. On Reddit tasks from WebArena, ActionEngine achieves 95% task success with, on average, a single LLM call, compared to 66% for the strongest vision-only baseline. Cost drops by 11.8x. Latency drops by 2x. If the pre-planned script fails at runtime, a vision-based fallback repairs the action and updates the memory graph for future runs. Why does it matter? Treating GUI interaction as graph traversal rather than step-by-step probabilistic reasoning is a compelling direction for making agents both faster and more reliable. Paper: https://t.co/UR0PjvFf0c Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c

Media 1
Media 2

📊 Media Metadata

{
  "media": [
    {
      "type": "photo",
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2026678090815123594/media_0.jpg?",
      "filename": "media_0.jpg"
    },
    {
      "type": "photo",
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2026678090815123594/media_1.png?",
      "filename": "media_1.png"
    }
  ],
  "processed_at": "2026-03-01T19:15:46.754749",
  "pipeline_version": "2.0"
}

🔧 Raw API Response

{
  "type": "tweet",
  "id": "2026678090815123594",
  "url": "https://x.com/dair_ai/status/2026678090815123594",
  "twitterUrl": "https://twitter.com/dair_ai/status/2026678090815123594",
  "text": "New research from Georgia Tech and Microsoft Research.\n\nGUI agents today are reactive. Every step costs an LLM call, which is why a lot of GUI agents are expensive, slow, and fragile.\n\nThis new research introduces ActionEngine, a framework that shifts GUI agents from reactive execution to programmatic planning.\n\nA Crawling Agent explores the application offline and builds a state-machine graph of the interface. Nodes are page states, edges are actions. Then at runtime, an Execution Agent uses this graph to synthesize a complete Python program in a single LLM call.\n\nInstead of O(N) vision model calls per task, you get O(1) planning cost.\n\nOn Reddit tasks from WebArena, ActionEngine achieves 95% task success with, on average, a single LLM call, compared to 66% for the strongest vision-only baseline. Cost drops by 11.8x. Latency drops by 2x.\n\nIf the pre-planned script fails at runtime, a vision-based fallback repairs the action and updates the memory graph for future runs.\n\nWhy does it matter?\n\nTreating GUI interaction as graph traversal rather than step-by-step probabilistic reasoning is a compelling direction for making agents both faster and more reliable.\n\nPaper: https://t.co/UR0PjvFf0c\n\nLearn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c",
  "source": "Twitter for iPhone",
  "retweetCount": 36,
  "replyCount": 15,
  "likeCount": 250,
  "quoteCount": 2,
  "viewCount": 23474,
  "createdAt": "Wed Feb 25 15:18:05 +0000 2026",
  "lang": "en",
  "bookmarkCount": 241,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2026678090815123594",
  "displayTextRange": [
    0,
    276
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "dair_ai",
    "url": "https://x.com/dair_ai",
    "twitterUrl": "https://twitter.com/dair_ai",
    "id": "889050642903293953",
    "name": "DAIR.AI",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/1643277398522187778/31dedbLo_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/889050642903293953/1742055232",
    "description": "",
    "location": "",
    "followers": 90586,
    "following": 1,
    "status": "",
    "canDm": true,
    "canMediaTag": true,
    "createdAt": "Sun Jul 23 09:12:45 +0000 2017",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 4185,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 161,
    "statusesCount": 2963,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "2028094132090966088"
    ],
    "profile_bio": {
      "description": "Democratizing AI research, education, and technologies.",
      "entities": {
        "description": {
          "hashtags": [],
          "symbols": [],
          "urls": [],
          "user_mentions": []
        },
        "url": {
          "urls": [
            {
              "display_url": "dair.ai",
              "expanded_url": "https://www.dair.ai/",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/lkqPZtMmfU"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "display_url": "pic.twitter.com/0PnJZYYefu",
        "expanded_url": "https://twitter.com/dair_ai/status/2026678090815123594/photo/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "features": {
          "large": {
            "faces": []
          },
          "orig": {
            "faces": []
          }
        },
        "id_str": "2026678086788591616",
        "indices": [
          277,
          300
        ],
        "media_key": "3_2026678086788591616",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARwgNPttGjAACgACHCA0/F0aMIoAAA==",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAABCgABHCA0+20aMAAKAAIcIDT8XRowigAA",
            "media_key": "3_2026678086788591616"
          }
        },
        "media_url_https": "https://pbs.twimg.com/media/HCA0-20aMAAIdm6.jpg",
        "original_info": {
          "focus_rects": [
            {
              "h": 904,
              "w": 1614,
              "x": 0,
              "y": 0
            },
            {
              "h": 1614,
              "w": 1614,
              "x": 0,
              "y": 0
            },
            {
              "h": 1798,
              "w": 1577,
              "x": 0,
              "y": 0
            },
            {
              "h": 1798,
              "w": 899,
              "x": 315,
              "y": 0
            },
            {
              "h": 1798,
              "w": 1614,
              "x": 0,
              "y": 0
            }
          ],
          "height": 1798,
          "width": 1614
        },
        "sizes": {
          "large": {
            "h": 1798,
            "w": 1614
          }
        },
        "type": "photo",
        "url": "https://t.co/0PnJZYYefu"
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "symbols": [],
    "urls": [
      {
        "display_url": "arxiv.org/abs/2602.20502",
        "expanded_url": "https://arxiv.org/abs/2602.20502",
        "indices": [
          1183,
          1206
        ],
        "url": "https://t.co/UR0PjvFf0c"
      },
      {
        "display_url": "academy.dair.ai",
        "expanded_url": "https://academy.dair.ai/",
        "indices": [
          1259,
          1282
        ],
        "url": "https://t.co/LRnpZN7L4c"
      }
    ],
    "user_mentions": []
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "article": null
}