🐦 Twitter Post Details

Viewing enriched Twitter post

@Akashi203

i open-sourced autokernel -- autoresearch for GPU kernels you give it any pytorch model. it profiles the model, finds the bottleneck kernels, writes triton replacements, and runs experiments overnight. edit one file, benchmark, keep or revert, repeat forever. same loop as @karpathy autoresearch, applied to kernel optimization 95 experiments. 18 TFLOPS → 187 TFLOPS. 1.31x vs cuBLAS. all autonomous 9 kernel types (matmul, flash attention, fused mlp, layernorm, rmsnorm, softmax, rope, cross entropy, reduce). amdahl's law decides what to optimize next. 5-stage correctness checks before any speedup counts the agent reads program.md (the "research org code"), edits https://t.co/ubU2WPOJwh, runs https://t.co/gJklhqV5hS, and either keeps or reverts. ~40 experiments/hour. ~320 overnight ships with self-contained GPT-2, LLaMA, and BERT definitions so you don't need the transformers library to get started https://t.co/45z8Z7nP3N

Media 1
Media 2

📊 Media Metadata

{
  "media": [
    {
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2031533857082646769/media_0.jpg",
      "media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2031533857082646769/media_0.jpg",
      "type": "photo",
      "filename": "media_0.jpg"
    },
    {
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2031533857082646769/media_1.jpg",
      "media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2031533857082646769/media_1.jpg",
      "type": "photo",
      "filename": "media_1.jpg"
    }
  ],
  "processed_at": "2026-03-12T14:03:06.846178",
  "pipeline_version": "2.0"
}

🔧 Raw API Response

{
  "type": "tweet",
  "id": "2031533857082646769",
  "url": "https://x.com/Akashi203/status/2031533857082646769",
  "twitterUrl": "https://twitter.com/Akashi203/status/2031533857082646769",
  "text": "i open-sourced autokernel -- autoresearch for GPU kernels\n\nyou give it any pytorch model. it profiles the model, finds the bottleneck kernels, writes triton replacements, and runs experiments overnight. edit one file, benchmark, keep or revert, repeat forever.\n\nsame loop as @karpathy autoresearch, applied to kernel optimization\n\n95 experiments. 18 TFLOPS → 187 TFLOPS. 1.31x vs cuBLAS. all autonomous\n\n9 kernel types (matmul, flash attention, fused mlp, layernorm, rmsnorm, softmax, rope, cross entropy, reduce). amdahl's law decides what to optimize next. 5-stage correctness checks before any speedup counts\n\nthe agent reads program.md (the \"research org code\"), edits https://t.co/ubU2WPOJwh, runs https://t.co/gJklhqV5hS, and either keeps or reverts. ~40 experiments/hour. ~320 overnight\n\nships with self-contained GPT-2, LLaMA, and BERT definitions so you don't need the transformers library to get started\n\nhttps://t.co/45z8Z7nP3N",
  "source": "Twitter for iPhone",
  "retweetCount": 165,
  "replyCount": 57,
  "likeCount": 1643,
  "quoteCount": 36,
  "viewCount": 171792,
  "createdAt": "Wed Mar 11 00:53:10 +0000 2026",
  "lang": "en",
  "bookmarkCount": 1342,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2031533857082646769",
  "displayTextRange": [
    0,
    274
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "Akashi203",
    "url": "https://x.com/Akashi203",
    "twitterUrl": "https://twitter.com/Akashi203",
    "id": "1826605998699659265",
    "name": "Jaber",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/2015781426981916672/D-gDMIO7_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/1826605998699659265/1756331997",
    "description": "founder @rightnowai_co | building the agi compiler on my gtx 1060",
    "location": "Hashemite Kingdom of Jordan",
    "followers": 3313,
    "following": 302,
    "status": "",
    "canDm": true,
    "canMediaTag": true,
    "createdAt": "Thu Aug 22 13:03:12 +0000 2024",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {
        "urls": [
          {
            "display_url": "rightnowai.co",
            "expanded_url": "http://rightnowai.co",
            "indices": [
              0,
              23
            ],
            "url": "https://t.co/Ze9HuduDku"
          }
        ]
      }
    },
    "fastFollowersCount": 0,
    "favouritesCount": 949,
    "hasCustomTimelines": false,
    "isTranslator": false,
    "mediaCount": 73,
    "statusesCount": 384,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "2016085729802080338"
    ],
    "profile_bio": {},
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "allow_download_status": {
          "allow_download": true
        },
        "display_url": "pic.x.com/LhOZCQZRhY",
        "expanded_url": "https://x.com/Akashi203/status/2031533857082646769/photo/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "features": {
          "large": {
            "faces": []
          },
          "medium": {
            "faces": []
          },
          "orig": {
            "faces": []
          },
          "small": {
            "faces": []
          }
        },
        "id_str": "2031532748599603205",
        "indices": [
          275,
          298
        ],
        "media_key": "3_2031532748599603205",
        "media_results": {
          "result": {
            "media_key": "3_2031532748599603205"
          }
        },
        "media_url_https": "https://pbs.twimg.com/media/HDF0RXXaMAUpkJT.jpg",
        "original_info": {
          "focus_rects": [
            {
              "h": 1236,
              "w": 2208,
              "x": 0,
              "y": 0
            },
            {
              "h": 1664,
              "w": 1664,
              "x": 272,
              "y": 0
            },
            {
              "h": 1664,
              "w": 1460,
              "x": 374,
              "y": 0
            },
            {
              "h": 1664,
              "w": 832,
              "x": 688,
              "y": 0
            },
            {
              "h": 1664,
              "w": 2208,
              "x": 0,
              "y": 0
            }
          ],
          "height": 1664,
          "width": 2208
        },
        "sizes": {
          "large": {
            "h": 1543,
            "resize": "fit",
            "w": 2048
          },
          "medium": {
            "h": 904,
            "resize": "fit",
            "w": 1200
          },
          "small": {
            "h": 512,
            "resize": "fit",
            "w": 680
          },
          "thumb": {
            "h": 150,
            "resize": "crop",
            "w": 150
          }
        },
        "type": "photo",
        "url": "https://t.co/LhOZCQZRhY"
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "symbols": [],
    "urls": [
      {
        "display_url": "kernel.py",
        "expanded_url": "http://kernel.py",
        "indices": [
          673,
          696
        ],
        "url": "https://t.co/ubU2WPOJwh"
      },
      {
        "display_url": "bench.py",
        "expanded_url": "http://bench.py",
        "indices": [
          703,
          726
        ],
        "url": "https://t.co/gJklhqV5hS"
      },
      {
        "display_url": "github.com/RightNow-AI/au…",
        "expanded_url": "http://github.com/RightNow-AI/autokernel",
        "indices": [
          915,
          938
        ],
        "url": "https://t.co/45z8Z7nP3N"
      }
    ],
    "user_mentions": [
      {
        "id_str": "33836629",
        "indices": [
          275,
          284
        ],
        "name": "Andrej Karpathy",
        "screen_name": "karpathy"
      }
    ]
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "article": null
}