{
  "type": "tweet",
  "id": "2032109644362223679",
  "url": "https://x.com/bytedunks/status/2032109644362223679",
  "twitterUrl": "https://twitter.com/bytedunks/status/2032109644362223679",
  "text": "Lately, I have been thinking about continual learning --- could this unlock novel interfaces to solve more complex problems with AI?\n\nAn experiment: Self-Improving Browser Agents (SIBA)\n\nBrowser agents are currently difficult to steer just by prompting, making them hardly capable at executing workflows reliably and flexibly without parametric updates.\n\nI could never get actual tasks done with Browser Agents previously, and thought this could be a worthy attempt at changing this.\n\nThe demo shows the SIBA agent extracting 3 receipts from my Gmail, and saving them into a local directory, tagged by software, for what would be a monthly software claim submission process.\n\nWhat differentiates SIBA from previous methods:\n\n1) CLI-driven browser control\nFollowing agent-browser, we select web components by textual references which significantly reduces incorrect component selection.\n\nFurthermore, LLMs work best reasoning about code. Previous methods rely heavily on a model's visual understanding to select the correct visual component by generating the (x,y) coordinates of a browser window (which are expected to be almost precise).\n\nThe emerging \"Skills + CLI\" design pattern for agentic tool calling will be included in the next iteration.\n\n2) Meta-agent driven improvement\nTuning a browser agent's system by hand is ineffective: the root cause of most failure modes are ambiguous and difficult to address systematically without parametric updates.\n\nMeta-agent driven improvement based on human feedback show promising initial results for now --- it took me three iterations of feedback and some API credits to get my custom workflow working.\n\nMoreover, providing clear feedback is something humans are good at, and also a user experience that feels natural.\n\nCaveat: currently, a simple meta-agent is used for iterative improvement. There are more effective ways to  do so systematically that will be explored next.\n\nMy hope is to use SIBA for smoke testing at scale. It would be interesting to see if it could be extended to persona-based smoke testing using parallel agents. Let's see.",
  "source": "Twitter for iPhone",
  "retweetCount": 1,
  "replyCount": 1,
  "likeCount": 1,
  "quoteCount": 0,
  "viewCount": 24,
  "createdAt": "Thu Mar 12 15:01:08 +0000 2026",
  "lang": "en",
  "bookmarkCount": 0,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2032109644362223679",
  "displayTextRange": [
    0,
    280
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "bytedunks",
    "url": "https://x.com/bytedunks",
    "twitterUrl": "https://twitter.com/bytedunks",
    "id": "993931267744841729",
    "name": "Brandon Ong",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/1924820134822379520/VS9gr799_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/993931267744841729/1692608929",
    "description": "",
    "location": "SG",
    "followers": 411,
    "following": 248,
    "status": "",
    "canDm": false,
    "canMediaTag": true,
    "createdAt": "Tue May 08 19:10:55 +0000 2018",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 163,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 20,
    "statusesCount": 286,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "2026950810102616571"
    ],
    "profile_bio": {
      "description": "exploring;\nprev Joint PhD in Robotics (on leave) @Columbia @NTUsg",
      "entities": {
        "description": {
          "hashtags": [],
          "symbols": [],
          "urls": [],
          "user_mentions": [
            {
              "id_str": "0",
              "indices": [
                49,
                58
              ],
              "name": "",
              "screen_name": "Columbia"
            },
            {
              "id_str": "0",
              "indices": [
                59,
                65
              ],
              "name": "",
              "screen_name": "NTUsg"
            }
          ]
        },
        "url": {
          "urls": [
            {
              "display_url": "ob1.info",
              "expanded_url": "https://ob1.info",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/tPv3PGYnpe"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "additional_media_info": {
          "monetizable": false
        },
        "display_url": "pic.twitter.com/VFj7C7Qszg",
        "expanded_url": "https://twitter.com/bytedunks/status/2032109644362223679/video/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "id_str": "2032105651909079040",
        "indices": [
          281,
          304
        ],
        "media_key": "13_2032105651909079040",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwABAoAARwzfVLoGrAAAAA=",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAAECgABHDN9UugasAAAAA==",
            "media_key": "13_2032105651909079040"
          }
        },
        "media_url_https": "https://pbs.twimg.com/amplify_video_thumb/2032105651909079040/img/itMOoReIFHAt2DNj.jpg",
        "original_info": {
          "focus_rects": [],
          "height": 720,
          "width": 1280
        },
        "sizes": {
          "large": {
            "h": 720,
            "w": 1280
          }
        },
        "type": "video",
        "url": "https://t.co/VFj7C7Qszg",
        "video_info": {
          "aspect_ratio": [
            16,
            9
          ],
          "duration_millis": 67633,
          "variants": [
            {
              "content_type": "application/x-mpegURL",
              "url": "https://video.twimg.com/amplify_video/2032105651909079040/pl/NNfAQvg_BaEMZwEV.m3u8?tag=21"
            },
            {
              "bitrate": 256000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2032105651909079040/vid/avc1/480x270/Ncw4x6dnepQu6817.mp4?tag=21"
            },
            {
              "bitrate": 832000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2032105651909079040/vid/avc1/640x360/sfTiqvz3a5ce3zH-.mp4?tag=21"
            },
            {
              "bitrate": 2176000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2032105651909079040/vid/avc1/1280x720/b44fFvy46xydAidz.mp4?tag=21"
            }
          ]
        }
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "symbols": [],
    "timestamps": [],
    "urls": [],
    "user_mentions": []
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "article": null
}