Hacker News: atgctg

GPT-5.2

atgctg — Thu, 11 Dec 2025 18:04:47 +0000

https://platform.openai.com/docs/guides/latest-model

System card: https://cdn.openai.com/pdf/3a4153c8-c748-4b71-8e31-aecbde944...

Comments URL: https://news.ycombinator.com/item?id=46234788

Points: 1195

# Comments: 1083

Llama.cpp: Add GPT-OSS

atgctg — Tue, 05 Aug 2025 15:23:58 +0000

Article URL: https://github.com/ggml-org/llama.cpp/pull/15091

Comments URL: https://news.ycombinator.com/item?id=44799218

Points: 35

# Comments: 0

New comment by atgctg in "Transformers Without Normalization"

atgctg — Sat, 15 Mar 2025 10:40:15 +0000

The paper's Table 7 shows DyT reducing overall LLaMA 7B inference time by 7.8% and training time by 8.2%. That is not insignificant.

New comment by atgctg in "An Interview with Daniel Gross and Nat Friedman About Models, Margins, and Moats"

atgctg — Thu, 23 Jan 2025 14:53:19 +0000

You can get a free trial right now to Stratechery Plus through Asianometry:

https://stratechery.passport.online/member/plan/4ycW4SE71Cy6...

Source: https://substack.com/home/post/p-154928959

I Love Calculator

atgctg — Sat, 28 Sep 2024 10:12:38 +0000

Article URL: https://karpathy.ai/blog/calculator.html

Comments URL: https://news.ycombinator.com/item?id=41679255

Points: 6

# Comments: 1

New comment by atgctg in "Prompt Caching"

atgctg — Mon, 19 Aug 2024 07:01:56 +0000

You have to store the KV cache, not the tokens. For Gemma 27B (probably slightly larger than Flash), this would be:

  Size of KV cache = 2 * (num_layers) * (num_kv_heads * dim_head) * seq_length * precision

  8-bit Gemma 27B KV cache = 2 * (46) * (16 * 144) * 1e6 * 1 byte ≈ 200 GB

Note that this doesn't take further optimizations into account that Google might be using.

Formula: https://developer.nvidia.com/blog/mastering-llm-techniques-i...

Gemma 27B config: https://huggingface.co/google/gemma-2-27b/blob/main/config.j...

Open source AI is the path forward

atgctg — Tue, 23 Jul 2024 15:08:41 +0000

Article URL: https://about.fb.com/news/2024/07/open-source-ai-is-the-path-forward/

Comments URL: https://news.ycombinator.com/item?id=41046773

Points: 2360

# Comments: 887

The Perceptron Controversy

atgctg — Fri, 14 Jun 2024 15:34:11 +0000

Article URL: https://yuxi-liu-wired.github.io/essays/posts/perceptron-controversy/

Comments URL: https://news.ycombinator.com/item?id=40681806

Points: 4

# Comments: 0

New comment by atgctg in "GitHub Profile with a Custom Background"

atgctg — Fri, 07 Jun 2024 19:24:00 +0000

Works using math CSS injection [1]:

    ```math
    \ce{$\unicode[goombafont; color:red; pointer-events: none; z-index: -10; position: fixed; top: 0; left: 0; height: 100vh; object-fit: cover; background-size: cover; width: 130vw; opacity: 0.5; background: url('https://github.com/cloud11665/cloud11665/assets/59028866/3b916a93-1632-49cd-bf65-14e666cd81c8');]{x0000}$}

[1]: https://raw.githubusercontent.com/cloud11665/cloud11665/mast...

New comment by atgctg in "GPT-4o"

atgctg — Mon, 13 May 2024 17:30:03 +0000

Tiktoken added support for GPT-4o: https://github.com/openai/tiktoken/commit/9d01e5670ff50eb74c...

It has an increased vocab size of 200k.

New comment by atgctg in "OpenAI: Model Spec"

atgctg — Wed, 08 May 2024 21:14:42 +0000

Seems like they are working on adding that capability:

> We're exploring whether we can responsibly provide the ability to generate NSFW content in age-appropriate contexts through the API and ChatGPT.

Link to section: https://cdn.openai.com/spec/model-spec-2024-05-08.html#dont-...

Stripe re-enters crypto with stablecoin payments

atgctg — Thu, 25 Apr 2024 17:32:52 +0000

Article URL: https://twitter.com/BackseatVC/status/1783544518320742774

Comments URL: https://news.ycombinator.com/item?id=40160572

Points: 3

# Comments: 0

Poor Charlie's Almanack (Berkshire Mode)

atgctg — Wed, 06 Dec 2023 07:22:23 +0000

Article URL: https://www.stripe.press/poor-charlies-almanack/book

Comments URL: https://news.ycombinator.com/item?id=38541282

Points: 3

# Comments: 0

New comment by atgctg in "LLM Visualization"

atgctg — Sun, 03 Dec 2023 18:34:44 +0000

A lot of transformer explanations fail to mention what makes self attention so powerful.

Unlike traditional neural networks with fixed weights, self-attention layers adaptively weight connections between inputs based on context. This allows transformers to accomplish in a single layer what would take traditional networks multiple layers.