Top all-time · #4382
Cameron R. Wolfe
Personal blog · cameronrwolfe.substack.comai
Cameron R. Wolfe writes Deep (Learning) Focus, a personal blog at cameronrwolfe.substack.com about ai. Since 2023, 18 different Hacker News readers have linked it 40 times, most often “Scaling Laws for LLMs: From GPT-3 to o3”. It is still writing: the last post was on 2026-10-21, and you can follow it by feed.
Too few sampled comments (7) to say how often HN recommends it.?
Why HN links it
HN readers link this Substack for deep dives on LLM topics: decoder-only transformers, finetuned LLM judges, and GRPO training tricks. It is also listed among AI news sources.
Summarised by Claude from a sample of 7 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
I contextualize and explain important topics in AI research. Click to read Deep (Learning) Focus, by Cameron R. Wolfe, Ph.D.,
Most-linked posts?
- Scaling Laws for LLMs: From GPT-3 to o33 people · 3 mentions · first linked 2025-01-08
- /3 people · 3 mentions · first linked 2023-11-12
- Mixture-of-Experts (Moe) LLMs2 people · 2 mentions · first linked 2025-01-27
- Decoder-Only Transformers: The Workhorse of Generative LLMs2 people · 2 mentions · first linked 2024-03-05
- Understanding LLaMA-2 from the Ground Up2 people · 2 mentions · first linked 2023-08-14
- Open-Source LLMs: Better Base Models (Part Two)2 people · 2 mentions · first linked 2023-07-31
What Hacker News says?
I think what you suggest would be very similar to a encoder-decoder architecture, which has been abandoned in favor of decoder-only architectures ( https://cameronrwolfe.substack.com/p/decoder-only-transforme... ). So I am guessing that what you suggest has already been tried and didn't work out, ut not sure why (the problems I mentioned above or something else). Sorry, that's where the limit of my knowledge is. I w…
Comment linking /p/decoder-only-transformers-the-workhorse
on “Better and Faster Large Language Models via Multi-Token Prediction”Submitted Olmo 3 and the Open LLM Renaissance as a story · 4 points
HN discussionToo late to edit, but here's a great, really in-depth post about using LLMs as judges to evaluate LLM outputs (when you don't have the ground truth for everything): https://cameronrwolfe.substack.com/p/finetuned-judge This is about finetuning LLMs to do it, but the first part is a good intro to why and how.
Comment linking /p/finetuned-judge
on “Test Driven Development (TDD) for your LLMs? Yes please, more of that please”Submitted NanoMoE: Mixture-of-Experts (Moe) LLMs from Scratch in PyTorch as a story · 4 points
HN discussionTheir R1 paper was really well-done. But I think it leaves out a few details necessary for stable training. https://cameronrwolfe.substack.com/p/grpo-tricks
Comment linking /p/grpo-tricks
on “DeepSeek open-sources inference optimizations with 60–85% faster generation [pdf]”Submitted Reinforcement Learning for LLMs: The Complete Guide as a story · 3 points
HN discussion> We don't have a model for a LLM that does exist will respond to a specific query either. Yes we do... It's math, you can calculate it. > No, it's not and I'm getting tired of explaining this. If you think it is, write your paper and get very rich. Why would I get rich for explaining how to do math? > You clearly don't understand any of this. Could you be more specific? Quantum physics is stupidly hard to calculate…
Comment linking /p/llm-scaling-laws
on “The Singularity will occur on a Tuesday”Submitted RL Scaling Laws for LLMs as a story · 3 points
HN discussionAI Supremacy: https://aisupremacy.substack.com/ Latent Space: https://www.latent.space/ Encyclopedia Autonomica: https://jdsemrau.substack.com/ Deep Learning Focus: https://cameronrwolfe.substack.com/ Artificial Fintelligence: https://finbarrtimbers.substack.com/ and I did not click on the links, but used a search engine to find the sources.
Comment linking /
on “Ask HN: Which sources cover AI developments without falling for the hype?”Submitted A Guide for Debugging LLM Training Data as a story · 3 points
HN discussionAI Supremacy: https://aisupremacy.substack.com/ Latent Space: https://www.latent.space/ Encyclopedia Autonomica: https://jdsemrau.substack.com/ Deep Learning Focus: https://cameronrwolfe.substack.com/ Artifical Fintelligence: https://finbarrtimbers.substack.com/
Comment linking /
on “Ask HN: Which sources cover AI developments without falling for the hype?”Submitted Vision Large Language Models (VLLMs) as a story · 3 points
HN discussionhttps://aisupremacy.substack.com/ https://www.latent.space/ https://jdsemrau.substack.com/ https://cameronrwolfe.substack.com/ https://finbarrtimbers.substack.com/
Comment linking /
on “Ask HN: Which sources cover AI developments without falling for the hype?”Submitted Open-Source LLMs: Better Base Models (Part Two) as a story · 3 points
HN discussionSubmitted Agent Evaluation: A Detailed Guide as a story · 2 points
HN discussionSubmitted Applying Statistics to LLM Evaluations as a story · 2 points
HN discussionSubmitted PPO for LLMs: A Guide for Normal People as a story · 2 points
HN discussionSubmitted AI Agents from First Principles as a story · 2 points
HN discussionSubmitted Demystifying Reasoning Models as a story · 2 points
HN discussionSubmitted Mixture-of-Experts (Moe) LLMs as a story · 2 points
HN discussionSubmitted Model Merging: A Survey as a story · 2 points
HN discussionSubmitted Decoder-Only Transformers: The Workhorse of Generative LLMs as a story · 2 points
HN discussionSubmitted The Basics of AI-Powered (Vector) Search as a story · 2 points
HN discussionSubmitted Open-Source LLMs: Imitation and Alignment (Part Three) as a story · 2 points
HN discussionSubmitted For those who've been missing out, here's the history of open source LLMs as a story · 2 points
HN discussionSubmitted Notes on Midtraining as a story · 1 points
HN discussionSubmitted Agentic World Models as a story · 1 points
HN discussionSubmitted The Anatomy of an LLM Benchmark as a story · 1 points
HN discussionSubmitted GPT-OSS from the Ground Up as a story · 1 points
HN discussionSubmitted Scaling Laws for LLMs: From GPT-3 to o3 as a story · 1 points
HN discussion
