I will teach you GPU by hand ✍️!

Date: 10/30

Time: 10am (Pacific Time)

Register: https://luma.com/khuz5upe

The leap from Hopper to Blackwell isn’t just a story of “faster GPUs” — it’s a fundamental rethinking of how every layer of the stack is organized to unlock the next frontier of AI scale. And as typical in an AI by Hand lecture, instead of focusing on benchmarks from the outside, we will dive inside the black box — unpacking the math behind these advances and making the underlying mechanisms accessible to you.

Outline

Primer: Grid -> Block -> Warp -> Thread
Quantization: FP8 -> MXFP4
Scaling: Layer -> Micro-Tensor
Packaging: Single-Die -> Dual-Die
Communication: In-GPU -> In-switch reduction

RSVP

This keynote lecture is sponsored by Together AI.

","contentEncoded","

I will teach you GPU by hand ✍️!

Date: 10/30

Time: 10am (Pacific Time)

Register: https://luma.com/khuz5upe

Outline

Primer: Grid -> Block -> Warp -> Thread
Quantization: FP8 -> MXFP4
Scaling: Layer -> Micro-Tensor
Packaging: Single-Die -> Dual-Die
Communication: In-GPU -> In-switch reduction

$\"\"$

RSVP

This keynote lecture is sponsored by Together AI.

","freeContentEncoded","textContent","I will teach you GPU by hand ✍️!Date: 10/30Time: 10am (Pacific Time)Register: https://luma.com/khuz5upeThe leap from Hopper to Blackwell isn’t just a story of “faster GPUs” — it’s a fundamental rethinking of how every layer of the stack is organized to unlock the next frontier of AI scale. And as typical in an AI by Hand lecture, instead of focusing on benchmarks from the outside, we will dive inside the black box — unpacking the math behind these advances and making the underlying mechanisms accessible to you.OutlinePrimer: Grid -> Block -> Warp -> ThreadQuantization: FP8 -> MXFP4Scaling: Layer -> Micro-TensorPackaging: Single-Die -> Dual-DieCommunication: In-GPU -> In-switch reductionRSVPThis keynote lecture is sponsored by Together AI.","published","strategy","content-ready","locked","deletedAt","2025-10-24T14:36:02.948Z","isSyncedFromOwnBlog","original","duplicatesCount","duplicateVersion","originalArticleId","errorsCount","appearance","color","backgroundColor","shortSummary","summary","bullets","persons","places","words","sentiment","subject","meta",[],"eventId","a4f18d9bc74af9d1cca73ca95258508b883a3f49e6ebacc9db5d1c01e2072f12","nostrRootEvent","event",{"_18":190,"_194":195,"_196":197,"_71":198,"_211":25,"_157":212,"_213":214},"sig","c75990366abc9edf9000f26992bbe93677d30ccece77dccc19fb89f04d705cf7c5225b1d6d5c9faa5f6817314dad73a2246b284824de015feaf042395e6241e3","kind",1,[199,201,202,206,209],[186,200],"GPU by Hand ✍️ Together",[34,149],[203,204,15,205],"a","30023:fa1756185f2879c4d47b882789bc588d3ad64030bf50c2f7e48eb57d2d8d02a8:f8acba97-723d-432f-ac13-b52254eb71e1","mention",[207,208],"client","Layer3.press",[210,25,15],"p","pubkey","GPU by Hand ✍️ Together\nKeynote Lecture Sponsored by Together AI\n\n\nnostr:naddr1qvzqqqr4gupzp7sh2cv972recn28hzp83x793rf66eqrp06sctm7fr4405kc6q4gqqjxvwrpvd3xzwfh95mnyvmy956rxvnx94skxvfn943r2v3jx56x2c3hx9jnz4klxae ","created_at",1761315861,"authorNPub","markdown","I will teach you GPU by hand ✍️!\n\n[](https://substackcdn.com/image/fetch/$s_!Upc8!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6602e63c-7d0d-489f-a7ee-9d37f40d87dd_3254x2148.jpeg)\n\n****Date:**** 10/30\n\n****Time:**** 10am (Pacific Time)\n\n****Register:**** [https://luma.com/khuz5upe](https://luma.com/khuz5upe \"https://luma.com/khuz5upe\")\n\nThe leap from ****Hopper**** to ****Blackwell**** isn’t just a story of “faster GPUs” — it’s a fundamental rethinking of how every layer of the stack is organized to unlock the next frontier of AI scale. And as typical in an __AI by Hand__ lecture, instead of focusing on benchmarks from the outside, we will dive inside the black box — unpacking the math behind these advances and making the underlying mechanisms accessible to you.\n\n****Outline****\n\n* Primer: Grid -> Block -> Warp -> Thread\n* Quantization: FP8 -> MXFP4\n* Scaling: Layer -> Micro-Tensor\n* Packaging: Single-Die -> Dual-Die\n* Communication: In-GPU -> In-switch reduction\n\n![](https://substackcdn.com/image/fetch/$s_!1z3v!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F5dfb94a7-4ca1-41ec-a735-c2b90facebc3_1242x582.png)\n\n[RSVP](https://luma.com/khuz5upe \"RSVP\")\n\nThis keynote lecture is sponsored by [Together AI](https://www.together.ai/ \"Together AI\").","freeMarkdown","kind1EventId","kind30023EventId","e6ed0c4768dbde6c6c7ddacc6b49b636e5eaf79ba004dba35d46b6db770b523c","kind30023",{"_18":221,"_194":224,"_196":225,"_71":226,"_211":25,"_157":246,"_213":247},"35bd23fd4e110fdd3a24669a362220739f5c33825c84a3318b09d9cdf01ea1c207ea6a8581046898620579362ba996bcef65e0aee146dc2d92070bfc2dced810",30023,[227,230,231,234,235,237,238,241,244,245],[228,229],"r","https://fakepeter.syndichain.com/articles/f8acba97-723d-432f-ac13-b52254eb71e1",[30,143],[232,233],"published_at","1760988020",[34,149],[236,138],"d",[180,144],[239,240],"prev-published-event-id","a1af47d31b9550ce26ac1fb312f6f3d0a545ba38a65d3800d9fa17b26d598105",[242,243],"revision","3",[207,208],[210,25,15],"GPU by Hand ✍️ Together (v3)\nI will teach you GPU by hand ✍️!\n\n[](https://substackcdn.com/image/fetch/$s_!Upc8!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6602e63c-7d0d-489f-a7ee-9d37f40d87dd_3254x2148.jpeg)\n\n****Date:**** 10/30\n\n****Time:**** 10am (Pacific Time)\n\n****Register:**** [https://luma.com/khuz5upe](https://luma.com/khuz5upe \"https://luma.com/khuz5upe\")\n\nThe leap from ****Hopper**** to ****Blackwell**** isn’t just a story of “faster GPUs” — it’s a fundamental rethinking of how every layer of the stack is organized to unlock the next frontier of AI scale. And as typical in an __AI by Hand__ lecture, instead of focusing on benchmarks from the outside, we will dive inside the black box — unpacking the math behind these advances and making the underlying mechanisms accessible to you.\n\n****Outline****\n\n* Primer: Grid -> Block -> Warp -> Thread\n* Quantization: FP8 -> MXFP4\n* Scaling: Layer -> Micro-Tensor\n* Packaging: Single-Die -> Dual-Die\n* Communication: In-GPU -> In-switch reduction\n\n![](https://substackcdn.com/image/fetch/$s_!1z3v!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F5dfb94a7-4ca1-41ec-a735-c2b90facebc3_1242x582.png)\n\n[RSVP](https://luma.com/khuz5upe \"RSVP\")\n\nThis keynote lecture is sponsored by [Together AI](https://www.together.ai/ \"Together AI\").\nOpen original article:\nhttps://www.byhand.ai/p/gpu-by-hand-together",1761316525,"publishedToNostr","htmlContentKind30023","

I will teach you GPU by hand ✍️!

Date: 10/30

Time: 10am (Pacific Time)

Register: https://luma.com/khuz5upe

The leap from Hopper to Blackwell isn’t just a story of “faster GPUs” — it’s a fundamental rethinking of how every layer of the stack is organized to unlock the next frontier of AI scale. And as typical in an AI by Hand lecture, instead of focusing on benchmarks from the outside, we will dive inside the black box — unpacking the math behind these advances and making the underlying mechanisms accessible to you.

Outline

Primer: Grid -> Block -> Warp -> Thread
Quantization: FP8 -> MXFP4
Scaling: Layer -> Micro-Tensor
Packaging: Single-Die -> Dual-Die
Communication: In-GPU -> In-switch reduction

$\"\"$

RSVP

This keynote lecture is sponsored by Together AI.

","pubDate","2025-10-20T19:20:20.000Z","modifiedAt","2025-10-23T10:10:09.948Z","2025-10-24T14:35:05.466Z","storyId","likes","reposts","comments","tweets",[],{"_18":19,"_20":21,"_22":23,"_24":25,"_26":-5,"_27":28,"_29":15,"_30":31,"_32":33,"_34":35,"_36":15,"_37":38,"_39":40,"_41":15,"_42":43,"_44":45,"_46":45,"_47":48,"_49":263,"_82":265},{"_18":48,"_51":23,"_52":23,"_53":54,"_32":33,"_55":56,"_57":-5,"_58":-5,"_59":35,"_60":61,"_62":-5,"_63":-5,"_64":65,"_66":67,"_68":25,"_69":70,"_71":264,"_73":-5,"_74":75,"_76":61,"_77":61,"_78":-5,"_79":-5,"_80":-5,"_81":-5,"_44":45,"_46":45},[],{"_84":19,"_85":-5,"_86":-5,"_87":88,"_89":90,"_91":266,"_94":-5,"_95":-5,"_12":13,"_96":267,"_98":268},{"_93":75},[],{},"relatedArticles",[],"actionData","errors"]