Top all-time · #169
Andrej Karpathy
Personal blog · karpathy.github.ioaiprogramming
Andrej Karpathy's personal blog at karpathy.github.io covers ai and programming. Since 2014, 263 different Hacker News readers have linked it 325 times, most often “The Unreasonable Effectiveness of Recurrent Neural Networks”. It is still writing: the last post was on 2026-02-12, and you can follow it by feed.
Sometimes recommended · 11 of 40 sampled comments recommend it?
Why HN links it
HN readers link Andrej Karpathy's blog for its accessible introductions to neural networks and RNNs ("The Unreasonable Effectiveness of RNNs"), reinforcement learning, and the "Hacker's guide" to neural nets. It is widely recommended as a starting point for ML.
- Recommended for · 11 comments
- Commenters call the RL post a great, plain-terms introduction that even non-practitioners can follow, the RNN post "excellent" and a good place to start, and sa
Summarised by Claude from a sample of 40 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Musings of a Computer Scientist.
Most-linked posts?
- The Unreasonable Effectiveness of Recurrent Neural Networks66 people · 77 mentions · first linked 2015-05-21
- The Unreasonable Effectiveness of Recurrent Neural Networks (2015)31 people · 31 mentions · first linked 2015-12-13
- Hacker's Guide to Neural Networks26 people · 29 mentions · first linked 2014-08-05
- The state of Computer Vision and AI: we are really, really far (2012)15 people · 16 mentions · first linked 2015-03-03
- What I learned from competing against a ConvNet on ImageNet (2014)13 people · 15 mentions · first linked 2014-09-04
- A Recipe for Training Neural Networks11 people · 11 mentions · first linked 2019-04-25
- A Recipe for Training Neural Networks10 people · 10 mentions · first linked 2019-10-10
- /2026/02/12/microgpt/10 people · 10 mentions · first linked 2026-02-16
- Deep Reinforcement Learning: Pong from Pixels10 people · 10 mentions · first linked 2016-05-31
- A from-scratch tour of Bitcoin in Python (2021)7 people · 8 mentions · first linked 2022-07-22
What Hacker News says?
I can't say what's happening in GitHub Copilot, but it's not necessarily true that the only way to produce syntactically valid outputs is to take substrings of the source text. It is possible to learn something approximating a generative grammar. Take a look at https://karpathy.github.io/2015/05/21/rnn-effectiveness/ At the same time, I would not be surprised if there are outputs that do correspond to the source tra…
Comment linking /2015/05/21/rnn-effectiveness/
on “GitHub Copilot”Submitted A from-scratch tour of Bitcoin in Python as a story · 1208 points
HN discussionWow, that post blew up. Like most engineers I respect JC for his incredible work, but I really think AGI is far off and at the moment would be very surprised if there is significant progress in the next years to come. I also want you to read this extremely well written (old) blog post about the topic and I don't think much has improved since: https://karpathy.github.io/2012/10/22/state-of-computer-visi...
Comment linking /2012/10/22/state-of-computer-vision/
on “John Carmack: I’m going to work on artificial general intelligence”Submitted The Unreasonable Effectiveness of Recurrent Neural Networks as a story · 913 points
HN discussionI very much enjoyed Andrej Karpathy‘s „from scratch“ bitcoin implementation [1]. I‘m sure there are other projects on GitHub explaining blockchain concepts directly in code. [1] https://karpathy.github.io/2021/06/21/blockchain/
Comment linking /2021/06/21/blockchain/
on “The Ethereum merge is done”Submitted Biohacking Lite as a story · 731 points
HN discussionOk, I'm talking about this claim in the top level post: > As an aside, odd to see people here are so pessimistic about the idea I suspect people in the comments are pessimistic about the idea because 1) they've seen a lot of AI hype [1] or 2) they don't know who karpathy is. His YouTube videos (which are tutorials about AI) consistently garner praise from "hacking-minded" people (the first post of his I read was hac…
Comment linking /neuralnets/
on “I am starting an AI+Education company”Submitted Hacker's Guide to Neural Networks as a story · 450 points
HN discussionRelationships between objects has been a problem with computer vision for a long time. 10 years ago: https://karpathy.github.io/2012/10/22/state-of-computer-visi... Now: https://arxiv.org/pdf/2204.13807 Given that this is what makes photos and videos interesting I think it's still a while before artists are automated.
Comment linking /2012/10/22/state-of-computer-vision/
on “Imagen Video: high definition video generation with diffusion models”Submitted Hacker's guide to Neural Networks (2012) as a story · 424 points
HN discussionIn his free time, I hope he writes some more fiction, I really liked https://karpathy.github.io/2015/11/14/ai/
Comment linking /2015/11/14/ai/
on “Hi everyone yes, I left OpenAI yesterday”Submitted A Recipe for Training Neural Networks as a story · 413 points
HN discussionIn 2015 he wrote this blog post about "The Unreasonable Effectiveness of Recurrent Neural Networks": https://karpathy.github.io/2015/05/21/rnn-effectiveness/ That blog post inspired Alec Radford at Open AI to do the research that produced the "Unsupervised sentiment neuron": https://openai.com/research/unsupervised-sentiment-neuron Open AI decided to see what happened if they scaled up that model by leveraging the n…
Comment linking /2015/05/21/rnn-effectiveness/
on “Hi everyone yes, I left OpenAI yesterday”Submitted Hacker's guide to Neural Networks (2012) as a story · 383 points
HN discussionPersonally I recommend Stanford CSI 231n http://cs231n.stanford.edu/ Its specifically geared towards visual recognition, but it starts with the basics of machine learning and moves on to feed forward nets and covnets and covers RNNs and attention towards the end. The assignments are a great set of jupyter notebooks that really get your hands on the material and you can find a number of peoples complete assignments o…
Comment linking /neuralnets/
on “Ask HN: Best way to get started with AI?”Submitted A Survival Guide to a PhD as a story · 319 points
HN discussionEvery beginner in Neural networks should probablystart with this and follow with Karpathy's http://karpathy.github.io/neuralnets and may be later on http://neuralnetworksanddeeplearning.com/
Comment linking /neuralnets
on “Neural Networks Demystified”Submitted What a Deep Neural Network thinks about selfies as a story · 262 points
HN discussionIs guru being used in some special sense here? Cuz Karpathy has been an AI luminary for at least 10 years. This post for instance taught RL to a whole generation https://karpathy.github.io/2016/05/31/rl/ .
Comment linking /2016/05/31/rl/
on “Doing well in your courses: Andrej's advice for success (2013)”Submitted Hacker's guide to Neural Networks as a story · 236 points
HN discussionIs it wrong to be critical of research? Back in my previous life of doing basic research I scrutinized papers left and right. http://karpathy.github.io/2015/05/21/rnn-effectiveness/ towards the end has similar methodology and is 1.5 years old. Hype is an interesting thing especially when it comes from laymen.
Comment linking /2015/05/21/rnn-effectiveness/
on “Unsupervised sentiment neuron”Submitted The state of Computer Vision and AI: we are really, really far (2012) as a story · 190 points
HN discussionThat would probably work. Karpathy's character based RNN could detect semantic meaning in text and code. http://karpathy.github.io/2015/05/21/rnn-effectiveness/
Comment linking /2015/05/21/rnn-effectiveness/
on “Unsupervised sentiment neuron”Submitted Deep Reinforcement Learning: Pong from Pixels as a story · 189 points
HN discussionYes happy to share once the evaluation period is over, but sadly I dont think Im allowed to share beforehand. Will have to check when but I expect it should be in approximately 2 weeks. In the meantime, you might want to check out this excellent blog post http://r2rt.com/recurrent-neural-networks-in-tensorflow-ii.h... . This will provide you with skeleton code to implement a character level generative model (similar…
Comment linking /2015/05/21/rnn-effectiveness/
on “Oxford Deep NLP – An advanced course on natural language processing”Submitted The Unreasonable Effectiveness of Recurrent Neural Networks (2015) as a story · 161 points
HN discussionAs Karpaty said: 1. Become one with the data. The first step to training a neural net is to not touch any neural net code at all and instead begin by thoroughly inspecting your data. This step is critical. I like to spend copious amount of time (measured in units of hours) scanning through thousands of examples, understanding their distribution and looking for patterns. Luckily, your brain is pretty good at this. On…
Comment linking /2019/04/25/recipe/
on “A dumb reason computer vision apps aren’t working: Exif Orientation”Submitted Self-driving as a case study for AGI as a story · 144 points
HN discussionJust looking at data and descriptive statistics is one of the first things a person is taught in machine learning, data science and statistics coursework. It’s a major skill in the field that is emphasized all the time. Practitioners frequently do cursory data analysis and data exploration to gain insight into the data, corner cases and which modeling approaches are plausible. Just to give some examples, Bayesian Da…
Comment linking /2019/04/25/recipe/
on “A dumb reason computer vision apps aren’t working: Exif Orientation”Submitted Deep Reinforcement Learning: Pong from Pixels (2016) as a story · 111 points
HN discussion
