Viewing enriched Twitter post
LLM APIs are not complete document parsers There’s been a lot of interest on how LLMs can replace existing OCR solutions altogether, and a baseline is screenshotting the page and feeding it to your favorite LLM API (OpenAI/Sonnet/Gemini). My argument is that 1) yes more… https://t.co/7ifO3veust
{
"media": [
{
"type": "image",
"url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/1945172322832539793/media_0.jpg?",
"filename": "media_0.jpg"
}
],
"nlp": {
"sentiment": "negative",
"topics": [
"LLM APIs",
"OCR solutions",
"document parsing"
],
"entities": [
"OpenAI",
"Sonnet",
"Gemini"
],
"summary": "The tweet argues that LLM APIs are not sufficient as complete document parsers compared to existing OCR solutions.",
"language": "en",
"processed_at": "2025-08-07T09:14:25.412310"
}
}{
"data": {
"threaded_conversation_with_injections_v2": {
"instructions": [
{
"type": "TimelineClearCache"
},
{
"type": "TimelineAddEntries",
"entries": [
{
"entryId": "tweet-1945172322832539793",
"sortIndex": "1953474806070378496",
"content": {
"entryType": "TimelineTimelineItem",
"__typename": "TimelineTimelineItem",
"itemContent": {
"itemType": "TimelineTweet",
"__typename": "TimelineTweet",
"tweet_results": {
"result": {
"__typename": "Tweet",
"rest_id": "1945172322832539793",
"has_birdwatch_notes": false,
"core": {
"user_results": {
"result": {
"__typename": "User",
"id": "VXNlcjozNjk3Nzc0MTY=",
"rest_id": "369777416",
"affiliates_highlighted_label": {},
"has_graduated_access": true,
"is_blue_verified": true,
"profile_image_shape": "Circle",
"legacy": {
"can_dm": true,
"can_media_tag": true,
"created_at": "Wed Sep 07 22:54:31 +0000 2011",
"default_profile": true,
"default_profile_image": false,
"description": "co-founder/CEO @llama_index\n\nCareers: https://t.co/EUnMNmbCtx\nEnterprise: https://t.co/Ht5jwxSrQB",
"entities": {
"description": {
"urls": [
{
"display_url": "llamaindex.ai/careers",
"expanded_url": "https://www.llamaindex.ai/careers",
"url": "https://t.co/EUnMNmbCtx",
"indices": [
38,
61
]
},
{
"display_url": "llamaindex.ai/contact",
"expanded_url": "https://www.llamaindex.ai/contact",
"url": "https://t.co/Ht5jwxSrQB",
"indices": [
74,
97
]
}
]
},
"url": {
"urls": [
{
"display_url": "llamaindex.ai",
"expanded_url": "https://www.llamaindex.ai/",
"url": "https://t.co/YiIfjVlzb6",
"indices": [
0,
23
]
}
]
}
},
"fast_followers_count": 0,
"favourites_count": 7776,
"followers_count": 64131,
"friends_count": 1378,
"has_custom_timelines": true,
"is_translator": false,
"listed_count": 1295,
"location": "",
"media_count": 1244,
"name": "Jerry Liu",
"normal_followers_count": 64131,
"pinned_tweet_ids_str": [
"1953108641558540720"
],
"possibly_sensitive": false,
"profile_image_url_https": "https://pbs.twimg.com/profile_images/1283610285031460864/1Q4zYhtb_normal.jpg",
"profile_interstitial_type": "",
"screen_name": "jerryjliu0",
"statuses_count": 6037,
"translator_type": "none",
"url": "https://t.co/YiIfjVlzb6",
"verified": false,
"want_retweets": false,
"withheld_in_countries": []
},
"tipjar_settings": {}
}
}
},
"unmention_data": {},
"edit_control": {
"edit_tweet_ids": [
"1945172322832539793"
],
"editable_until_msecs": "1752603795000",
"is_edit_eligible": true,
"edits_remaining": "5"
},
"is_translatable": false,
"views": {
"count": "6364",
"state": "EnabledWithCount"
},
"source": "<a href=\"https://mobile.twitter.com\" rel=\"nofollow\">Twitter Web App</a>",
"note_tweet": {
"is_expandable": true,
"note_tweet_results": {
"result": {
"id": "Tm90ZVR3ZWV0OjE5NDUxNzIzMjI2NDc5OTAyNzI=",
"text": "LLM APIs are not complete document parsers \n\nThere’s been a lot of interest on how LLMs can replace existing OCR solutions altogether, and a baseline is screenshotting the page and feeding it to your favorite LLM API (OpenAI/Sonnet/Gemini).\n\nMy argument is that 1) yes more traditional OCR solutions are dead because LLMs are really good, but 2) relying on LLM APIs on their own will not get you a complete parsing solution that you can deploy across every user.\n\nHere’s why:\n💡 There’s a lot of table-stakes metadata that will be missing. This includes confidence scores - needed if you need a human-in-the-loop step to validate the confidence of a page. This also includes bounding boxes, text matches, and reasoning - needed for citations, and fine-grained segmentation information (e.g. separating out paragraphs/charts) for chunking and other heuristic processing.\n\n💡There’s still a long-tail of accuracy issues. Screenshotting the page oftentimes misses out on relevant information directly found within the file binary. Even though models are getting much better (e.g. Gemin 2.5, Sonnet 4.0), there’s still areas where they fail.\n\n💡You need to standardize the prompts - there’s a big opportunity cost. Any LLM is only as good as the prompt you give it. Inevitably as you handle more document types, you are going to need to tweak the prompt/architecture and standardize it so it works across more doc types, and every team can use it. At that point you’re basically building/maintaining your own parsing solution.\n\n💡 Misc issues: Rate-limits, content filters, and reliability issues (e.g. when LLM is down) are all too common. Requires human effort to help maintain.\n\nOur point of view around LlamaCloud, our document processing solution, to wrap the latest LLM APIs so you get all their capabilities, but add on all the nice features above so you don’t have to build/maintain it yourself!\n\nCome check it out: https://t.co/XBj94oei28",
"entity_set": {
"hashtags": [],
"symbols": [],
"urls": [
{
"display_url": "cloud.llamaindex.ai/?utm_campaign=…",
"expanded_url": "https://cloud.llamaindex.ai/?utm_campaign=parse&utm_medium=jl_socials",
"url": "https://t.co/XBj94oei28",
"indices": [
1916,
1939
]
}
],
"user_mentions": []
},
"richtext": {
"richtext_tags": [
{
"from_index": 479,
"to_index": 539,
"richtext_types": [
"Bold"
]
},
{
"from_index": 873,
"to_index": 919,
"richtext_types": [
"Bold"
]
},
{
"from_index": 1141,
"to_index": 1211,
"richtext_types": [
"Bold"
]
},
{
"from_index": 1527,
"to_index": 1540,
"richtext_types": [
"Bold"
]
}
]
},
"media": {
"inline_media": []
}
}
}
},
"legacy": {
"bookmark_count": 37,
"bookmarked": false,
"created_at": "Tue Jul 15 17:23:15 +0000 2025",
"conversation_id_str": "1945172322832539793",
"display_text_range": [
0,
274
],
"entities": {
"hashtags": [],
"media": [
{
"display_url": "pic.x.com/7ifO3veust",
"expanded_url": "https://x.com/jerryjliu0/status/1945172322832539793/photo/1",
"id_str": "1945171980338032640",
"indices": [
275,
298
],
"media_key": "3_1945171980338032640",
"media_url_https": "https://pbs.twimg.com/media/Gv6jnDMXcAAR3BI.jpg",
"type": "photo",
"url": "https://t.co/7ifO3veust",
"ext_media_availability": {
"status": "Available"
},
"features": {
"large": {
"faces": []
},
"medium": {
"faces": []
},
"small": {
"faces": []
},
"orig": {
"faces": []
}
},
"sizes": {
"large": {
"h": 1035,
"w": 1351,
"resize": "fit"
},
"medium": {
"h": 919,
"w": 1200,
"resize": "fit"
},
"small": {
"h": 521,
"w": 680,
"resize": "fit"
},
"thumb": {
"h": 150,
"w": 150,
"resize": "crop"
}
},
"original_info": {
"height": 1035,
"width": 1351,
"focus_rects": [
{
"x": 0,
"y": 128,
"w": 1351,
"h": 757
},
{
"x": 192,
"y": 0,
"w": 1035,
"h": 1035
},
{
"x": 255,
"y": 0,
"w": 908,
"h": 1035
},
{
"x": 450,
"y": 0,
"w": 518,
"h": 1035
},
{
"x": 0,
"y": 0,
"w": 1351,
"h": 1035
}
]
},
"media_results": {
"result": {
"media_key": "3_1945171980338032640"
}
}
}
],
"symbols": [],
"timestamps": [],
"urls": [],
"user_mentions": []
},
"extended_entities": {
"media": [
{
"display_url": "pic.x.com/7ifO3veust",
"expanded_url": "https://x.com/jerryjliu0/status/1945172322832539793/photo/1",
"id_str": "1945171980338032640",
"indices": [
275,
298
],
"media_key": "3_1945171980338032640",
"media_url_https": "https://pbs.twimg.com/media/Gv6jnDMXcAAR3BI.jpg",
"type": "photo",
"url": "https://t.co/7ifO3veust",
"ext_media_availability": {
"status": "Available"
},
"features": {
"large": {
"faces": []
},
"medium": {
"faces": []
},
"small": {
"faces": []
},
"orig": {
"faces": []
}
},
"sizes": {
"large": {
"h": 1035,
"w": 1351,
"resize": "fit"
},
"medium": {
"h": 919,
"w": 1200,
"resize": "fit"
},
"small": {
"h": 521,
"w": 680,
"resize": "fit"
},
"thumb": {
"h": 150,
"w": 150,
"resize": "crop"
}
},
"original_info": {
"height": 1035,
"width": 1351,
"focus_rects": [
{
"x": 0,
"y": 128,
"w": 1351,
"h": 757
},
{
"x": 192,
"y": 0,
"w": 1035,
"h": 1035
},
{
"x": 255,
"y": 0,
"w": 908,
"h": 1035
},
{
"x": 450,
"y": 0,
"w": 518,
"h": 1035
},
{
"x": 0,
"y": 0,
"w": 1351,
"h": 1035
}
]
},
"media_results": {
"result": {
"media_key": "3_1945171980338032640"
}
}
}
]
},
"favorite_count": 40,
"favorited": false,
"full_text": "LLM APIs are not complete document parsers \n\nThere’s been a lot of interest on how LLMs can replace existing OCR solutions altogether, and a baseline is screenshotting the page and feeding it to your favorite LLM API (OpenAI/Sonnet/Gemini).\n\nMy argument is that 1) yes more… https://t.co/7ifO3veust",
"is_quote_status": false,
"lang": "en",
"possibly_sensitive": false,
"possibly_sensitive_editable": true,
"quote_count": 1,
"reply_count": 2,
"retweet_count": 8,
"retweeted": false,
"user_id_str": "369777416",
"id_str": "1945172322832539793"
},
"quick_promote_eligibility": {
"eligibility": "IneligibleNotProfessional"
}
}
},
"tweetDisplayType": "Tweet"
},
"clientEventInfo": {
"component": "tweet",
"element": "tweet"
}
}
},
{
"entryId": "conversationthread-1945203587274928325",
"sortIndex": "1953474806070378486",
"content": {
"entryType": "TimelineTimelineModule",
"__typename": "TimelineTimelineModule",
"items": [
{
"entryId": "conversationthread-1945203587274928325-tweet-1945203587274928325",
"item": {
"itemContent": {
"itemType": "TimelineTweet",
"__typename": "TimelineTweet",
"tweet_results": {
"result": {
"__typename": "Tweet",
"rest_id": "1945203587274928325",
"has_birdwatch_notes": false,
"core": {
"user_results": {
"result": {
"__typename": "User",
"id": "VXNlcjoxNDk0NjEwNzk4NjY4MzU3Njc1",
"rest_id": "1494610798668357675",
"affiliates_highlighted_label": {},
"has_graduated_access": true,
"is_blue_verified": false,
"profile_image_shape": "Circle",
"legacy": {
"can_dm": false,
"can_media_tag": true,
"created_at": "Fri Feb 18 09:54:46 +0000 2022",
"default_profile": true,
"default_profile_image": false,
"description": "Building @troopyevents. Software dev. Renowned and feared chapati chef (only at home). Into🚴♂️👨🌾🐄🐪🤖",
"entities": {
"description": {
"urls": []
}
},
"fast_followers_count": 0,
"favourites_count": 4255,
"followers_count": 322,
"friends_count": 504,
"has_custom_timelines": true,
"is_translator": false,
"listed_count": 1,
"location": "🇰🇪🇺🇬📍Nairobi",
"media_count": 70,
"name": "ChepKurui",
"normal_followers_count": 322,
"pinned_tweet_ids_str": [
"1894419495843672230"
],
"possibly_sensitive": false,
"profile_banner_url": "https://pbs.twimg.com/profile_banners/1494610798668357675/1695236753",
"profile_image_url_https": "https://pbs.twimg.com/profile_images/1563650541590396934/yEeSTukW_normal.jpg",
"profile_interstitial_type": "",
"screen_name": "MkrJmm1",
"statuses_count": 1325,
"translator_type": "none",
"verified": false,
"want_retweets": false,
"withheld_in_countries": []
},
"tipjar_settings": {}
}
}
},
"unmention_data": {},
"edit_control": {
"edit_tweet_ids": [
"1945203587274928325"
],
"editable_until_msecs": "1752611249000",
"is_edit_eligible": false,
"edits_remaining": "5"
},
"is_translatable": false,
"views": {
"count": "22",
"state": "EnabledWithCount"
},
"source": "<a href=\"http://twitter.com/download/iphone\" rel=\"nofollow\">Twitter for iPhone</a>",
"legacy": {
"bookmark_count": 0,
"bookmarked": false,
"created_at": "Tue Jul 15 19:27:29 +0000 2025",
"conversation_id_str": "1945172322832539793",
"display_text_range": [
25,
82
],
"entities": {
"hashtags": [],
"symbols": [],
"timestamps": [],
"urls": [],
"user_mentions": [
{
"id_str": "369777416",
"name": "Jerry Liu",
"screen_name": "jerryjliu0",
"indices": [
0,
11
]
},
{
"id_str": "3245334878",
"name": "Jon McCoy",
"screen_name": "thejonmccoy",
"indices": [
12,
24
]
}
]
},
"favorite_count": 0,
"favorited": false,
"full_text": "@jerryjliu0 @thejonmccoy Great post. Im checking out more on RAG at the right time",
"in_reply_to_screen_name": "jerryjliu0",
"in_reply_to_status_id_str": "1945172322832539793",
"in_reply_to_user_id_str": "369777416",
"is_quote_status": false,
"lang": "en",
"quote_count": 0,
"reply_count": 0,
"retweet_count": 0,
"retweeted": false,
"user_id_str": "1494610798668357675",
"id_str": "1945203587274928325"
},
"quick_promote_eligibility": {
"eligibility": "IneligibleNotProfessional"
}
}
},
"tweetDisplayType": "Tweet"
},
"clientEventInfo": {
"component": "tweet",
"element": "tweet",
"details": {
"conversationDetails": {
"conversationSection": "HighQuality"
},
"timelinesDetails": {
"controllerData": "DAACDAAEDAABCgABAAAAAAAAAAEKAAIAAAAAAAAAAAAAAAA="
}
}
}
}
}
],
"metadata": {
"conversationMetadata": {
"allTweetIds": [
"1945203587274928325"
],
"enableDeduplication": true
}
},
"displayType": "VerticalConversation",
"clientEventInfo": {
"component": "tweet",
"details": {
"conversationDetails": {
"conversationSection": "HighQuality"
},
"timelinesDetails": {
"controllerData": "DAACDAAEDAABCgABAAAAAAAAAAEKAAIAAAAAAAAAAAAAAAA="
}
}
}
}
},
{
"entryId": "cursor-bottom-1953474806070378485",
"sortIndex": "1953474806070378485",
"content": {
"entryType": "TimelineTimelineCursor",
"__typename": "TimelineTimelineCursor",
"value": "DAAFCgABGxwi_HV___MLAAIAAAAwRW1QQzZ3QUFBZlEvZ0dKTjB2R3AvQUFBQUFJYS9zQmJRMVp3eFJyK28rdnhXdUNSCAADAAAAAgAA",
"cursorType": "Bottom"
}
}
]
},
{
"type": "TimelineTerminateTimeline",
"direction": "Top"
},
{
"type": "TimelineTerminateTimeline",
"direction": "Bottom"
}
],
"metadata": {
"scribeConfig": {
"page": "ranked_replies"
}
}
}
}
}