Viewing enriched Twitter post
If you give GPT-5.4 a raw dump of the GPT-2 weights and ask for a <5000 byte C program to inference it, GPT-5.4 succeeds in under 15 minutes! I remember working on a similar exercise to compare results against a proprietary model in a previous paper - it took days!
{
"score": 0.4,
"score_components": {
"author": 0.09,
"engagement": 0.0,
"quality": 0.1,
"source": 0.135,
"nlp": 0.05,
"recency": 0.025
},
"scored_at": "2026-03-07T07:26:24.302349",
"import_source": "api_import",
"source_tagged_at": "2026-03-07T07:26:24.302371",
"enriched": true,
"enriched_at": "2026-03-07T07:26:24.302374"
}{
"type": "tweet",
"id": "2030151018164670486",
"url": "https://x.com/markchen90/status/2030151018164670486",
"twitterUrl": "https://twitter.com/markchen90/status/2030151018164670486",
"text": "If you give GPT-5.4 a raw dump of the GPT-2 weights and ask for a <5000 byte C program to inference it, GPT-5.4 succeeds in under 15 minutes!\n\nI remember working on a similar exercise to compare results against a proprietary model in a previous paper - it took days!",
"source": "Twitter for iPhone",
"retweetCount": 25,
"replyCount": 23,
"likeCount": 478,
"quoteCount": 2,
"viewCount": 49390,
"createdAt": "Sat Mar 07 05:18:15 +0000 2026",
"lang": "en",
"bookmarkCount": 117,
"isReply": false,
"inReplyToId": null,
"conversationId": "2030151018164670486",
"displayTextRange": [
0,
269
],
"inReplyToUserId": null,
"inReplyToUsername": null,
"author": {
"type": "user",
"userName": "markchen90",
"url": "https://x.com/markchen90",
"twitterUrl": "https://twitter.com/markchen90",
"id": "1273327266957688832",
"name": "Mark Chen",
"isVerified": false,
"isBlueVerified": true,
"verifiedType": null,
"profilePicture": "https://pbs.twimg.com/profile_images/1273327644436656129/uYf1oNS5_normal.jpg",
"coverPicture": "",
"description": "",
"location": "",
"followers": 70196,
"following": 350,
"status": "",
"canDm": false,
"canMediaTag": true,
"createdAt": "Wed Jun 17 18:50:59 +0000 2020",
"entities": {
"description": {
"urls": []
},
"url": {}
},
"fastFollowersCount": 0,
"favouritesCount": 2067,
"hasCustomTimelines": true,
"isTranslator": false,
"mediaCount": 23,
"statusesCount": 413,
"withheldInCountries": [],
"affiliatesHighlightedLabel": {},
"possiblySensitive": false,
"pinnedTweetIds": [
"1968372340271862014"
],
"profile_bio": {
"description": "Chief Research Officer at @OpenAI. Coach for the USA IOI Team.",
"entities": {
"description": {
"hashtags": [],
"symbols": [],
"urls": [],
"user_mentions": [
{
"id_str": "0",
"indices": [
26,
33
],
"name": "",
"screen_name": "OpenAI"
}
]
}
}
},
"isAutomated": false,
"automatedBy": null
},
"extendedEntities": {},
"card": null,
"place": {},
"entities": {
"hashtags": [],
"symbols": [],
"timestamps": [],
"urls": [],
"user_mentions": []
},
"quoted_tweet": {
"type": "tweet",
"id": "2030000810894184808",
"url": "https://x.com/hansonwng/status/2030000810894184808",
"twitterUrl": "https://twitter.com/hansonwng/status/2030000810894184808",
"text": "https://t.co/HnHV79hNcv",
"source": "Twitter for iPhone",
"retweetCount": 25,
"replyCount": 6,
"likeCount": 190,
"quoteCount": 14,
"viewCount": 121443,
"createdAt": "Fri Mar 06 19:21:23 +0000 2026",
"lang": "zxx",
"bookmarkCount": 154,
"isReply": false,
"inReplyToId": null,
"conversationId": "2030000810894184808",
"displayTextRange": [
0,
23
],
"inReplyToUserId": null,
"inReplyToUsername": null,
"author": {
"type": "user",
"userName": "hansonwng",
"url": "https://x.com/hansonwng",
"twitterUrl": "https://twitter.com/hansonwng",
"id": "1733089158",
"name": "Hanson Wang",
"isVerified": false,
"isBlueVerified": true,
"verifiedType": null,
"profilePicture": "https://pbs.twimg.com/profile_images/2020000268927262720/z0zRYz32_normal.jpg",
"coverPicture": "",
"description": "",
"location": "SF Bay Area",
"followers": 1312,
"following": 318,
"status": "",
"canDm": false,
"canMediaTag": true,
"createdAt": "Thu Sep 05 21:17:58 +0000 2013",
"entities": {
"description": {
"urls": []
},
"url": {}
},
"fastFollowersCount": 0,
"favouritesCount": 1446,
"hasCustomTimelines": true,
"isTranslator": false,
"mediaCount": 10,
"statusesCount": 101,
"withheldInCountries": [],
"affiliatesHighlightedLabel": {},
"possiblySensitive": false,
"pinnedTweetIds": [],
"profile_bio": {
"description": "@OpenAI Codex research // previously cofounder @arcwisedata",
"entities": {
"description": {
"hashtags": [],
"symbols": [],
"urls": [],
"user_mentions": [
{
"id_str": "0",
"indices": [
0,
7
],
"name": "",
"screen_name": "OpenAI"
},
{
"id_str": "0",
"indices": [
47,
59
],
"name": "",
"screen_name": "arcwisedata"
}
]
}
}
},
"isAutomated": false,
"automatedBy": null
},
"extendedEntities": {},
"card": null,
"place": {},
"entities": {
"hashtags": [],
"symbols": [],
"timestamps": [],
"urls": [
{
"display_url": "x.com/i/article/2029…",
"expanded_url": "http://x.com/i/article/2029955744620236801",
"indices": [
0,
23
],
"url": "https://t.co/HnHV79hNcv"
}
],
"user_mentions": []
},
"quoted_tweet": null,
"retweeted_tweet": null,
"isLimitedReply": false,
"article": {
"title": "GPT-5.4 code-golfs GPT-2",
"preview_text": "GPT-5.4 (along with its predecessor, GPT-5.3-Codex) are the first models to be able to consistently† solve the gpt2-codegolf problem in TerminalBench 2.0! It's a superhuman challenge where the model",
"cover_media_img_url": "https://pbs.twimg.com/media/HCvk7rPacAEuAcy.jpg"
}
},
"retweeted_tweet": null,
"isLimitedReply": false,
"article": null
}