🐦 Twitter Post Details

Viewing enriched Twitter post

@karpathy

@andrew_n_carr Yeah, $10B is the difference in finding it first and ~5 years ago. :) I just love reproducing landmark results for much cheaper, it's so fun! Reproducing LeCun 1989 was super fun too: https://t.co/oOZcQW3Y9H What runs unoptimized on a consumer laptop in 1 minute was a state of the art neural net trained for days in 1989. Another favorite example: CIFAR-10. In 2011 state of the art was 77%. I estimated human accuracy to be ~94% but said that performance might go up to 85-90%. https://t.co/KJl0V4T0ei Now you can speedrun to 94% accuracy in 1.98 seconds on a single GPU (yes, <2 seconds). https://t.co/wHUQs6htdV So e.g. right now GPT-2 (imo the landmark result that launched LLMs and where the modern stack is basically in full form) is ~$500, but I'm unreasonably obsessed with how much that can be brought down.

Media 1
View on Twitter

📊 Media Metadata

{
  "media": [
    {
      "type": "photo",
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2009406042267136238/media_0.jpg?",
      "filename": "media_0.jpg"
    }
  ],
  "processed_at": "2026-01-18T17:51:44.168136",
  "pipeline_version": "2.0"
}

🔧 Raw API Response

{
  "type": "tweet",
  "id": "2009406042267136238",
  "url": "https://x.com/karpathy/status/2009406042267136238",
  "twitterUrl": "https://twitter.com/karpathy/status/2009406042267136238",
  "text": "@andrew_n_carr Yeah, $10B is the difference in finding it first and ~5 years ago. :)\n\nI just love reproducing landmark results for much cheaper, it's so fun!\n\nReproducing LeCun 1989 was super fun too:\nhttps://t.co/oOZcQW3Y9H\nWhat runs unoptimized on a consumer laptop in 1 minute was a state of the art neural net trained for days in 1989.\n\nAnother favorite example: CIFAR-10. In 2011 state of the art was 77%. I estimated human accuracy to be ~94% but said that performance might go up to 85-90%.\nhttps://t.co/KJl0V4T0ei\nNow you can speedrun to 94% accuracy in 1.98 seconds on a single GPU (yes, <2 seconds).\nhttps://t.co/wHUQs6htdV\n\nSo e.g. right now GPT-2 (imo the landmark result that launched LLMs and where the modern stack is basically in full form) is ~$500, but I'm unreasonably obsessed with how much that can be brought down.",
  "source": "Twitter for iPhone",
  "retweetCount": 35,
  "replyCount": 18,
  "likeCount": 997,
  "quoteCount": 2,
  "viewCount": 60869,
  "createdAt": "Thu Jan 08 23:25:08 +0000 2026",
  "lang": "en",
  "bookmarkCount": 326,
  "isReply": true,
  "inReplyToId": "2009403497016668560",
  "conversationId": "2009403497016668560",
  "displayTextRange": [
    15,
    319
  ],
  "inReplyToUserId": "3378986176",
  "inReplyToUsername": "andrew_n_carr",
  "author": {
    "type": "user",
    "userName": "karpathy",
    "url": "https://x.com/karpathy",
    "twitterUrl": "https://twitter.com/karpathy",
    "id": "33836629",
    "name": "Andrej Karpathy",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/1296667294148382721/9Pr6XrPB_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/33836629/1407117611",
    "description": "",
    "location": "Stanford",
    "followers": 1609410,
    "following": 1048,
    "status": "",
    "canDm": true,
    "canMediaTag": true,
    "createdAt": "Tue Apr 21 06:49:15 +0000 2009",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 21607,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 841,
    "statusesCount": 9901,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "1617979122625712128"
    ],
    "profile_bio": {
      "description": "Building @EurekaLabsAI. Previously Director of AI @ Tesla, founding team @ OpenAI, CS231n/PhD @ Stanford. I like to train large deep neural nets.",
      "entities": {
        "description": {
          "user_mentions": [
            {
              "id_str": "0",
              "indices": [
                9,
                22
              ],
              "name": "",
              "screen_name": "EurekaLabsAI"
            }
          ]
        },
        "url": {
          "urls": [
            {
              "display_url": "karpathy.ai",
              "expanded_url": "https://karpathy.ai",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/0EcFthjJXM"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {},
  "card": null,
  "place": {},
  "entities": {
    "urls": [
      {
        "display_url": "karpathy.github.io/2022/03/14/lec…",
        "expanded_url": "https://karpathy.github.io/2022/03/14/lecun1989/",
        "indices": [
          186,
          209
        ],
        "url": "https://t.co/oOZcQW3Y9H"
      },
      {
        "display_url": "karpathy.github.io/2011/04/27/man…",
        "expanded_url": "https://karpathy.github.io/2011/04/27/manually-classifying-cifar10/",
        "indices": [
          483,
          506
        ],
        "url": "https://t.co/KJl0V4T0ei"
      },
      {
        "display_url": "github.com/hiverge/cifar1…",
        "expanded_url": "https://github.com/hiverge/cifar10-speedrun",
        "indices": [
          595,
          618
        ],
        "url": "https://t.co/wHUQs6htdV"
      }
    ]
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "article": null
}