2outube
Courses · Andrej Karpathy

Read every Karpathy Zero to Hero lecture as text

How do you get the transcripts of all of Karpathy Zero to Hero? Every lecture is listed below: open any one free (one video a day, no account), or get the whole series as one Markdown file, in order, ready to paste into your AI.

Get all 10 as one file 10 videos · one-time $4.99

The count and price are shown again before you pay. Videos with no caption track are marked and left out.

About this series

Andrej Karpathy's code-along series on building neural networks from scratch. It starts with backpropagation in micrograd, builds character-level language models in makemore, then constructs GPT, its tokenizer and a GPT-2 reproduction. The playlist has 10 videos.

Programmers following along who want to look up the exact explanation of a step, or ask an AI about a section without rewatching a 2-4 hour video.

All 10 lectures, in course order

  1. 1The spelled-out intro to neural networks and backpropagation: building micrograd2:25:52 · 2022-08-16
  2. 2The spelled-out intro to language modeling: building makemore1:57:45 · 2022-09-07
  3. 3Building makemore Part 2: MLP1:15:40 · 2022-09-12
  4. 4Building makemore Part 3: Activations & Gradients, BatchNorm1:55:58 · 2022-10-04
  5. 5Building makemore Part 4: Becoming a Backprop Ninja1:55:24 · 2022-10-11
  6. 6Building makemore Part 5: Building a WaveNet56:22 · 2022-11-21
  7. 7Let's build GPT: from scratch, in code, spelled out.1:56:20 · 2023-01-17
  8. 8State of GPT | BRK216HFS42:40 · 2023-05-25
  9. 9Let's build the GPT Tokenizer2:13:35 · 2024-02-20
  10. 10Let's reproduce GPT-2 (124M)4:01:26 · 2024-06-09

Watch on the creator's own pages: Andrej Karpathy on YouTube · the original playlist. 2outube is independent and not affiliated with Andrej Karpathy.

Ask your AI once it's in one file

  • Explain how micrograd implements backpropagation and the chain rule.
  • What problems does the course identify with initialization and how does batch normalization address them?
  • Walk through the transformer block as built in 'Let's build GPT'.
  • How does the tokenizer video explain byte-pair encoding and its quirks?
  • What training choices does the GPT-2 reproduction use for the optimizer, learning-rate schedule and batch size?

2outube gives you the text; your AI or notes app does the rest.

Questions, answered.

Can I get every Karpathy Zero to Hero video as text in one file?

Yes. Paste the playlist link into 2outube and it builds one Markdown file with all 10 videos in playlist order. The count and price are shown before you pay; this playlist fits the 25-video pack ($4.99), a one-time purchase. Videos with no caption track are marked and left out.

Can I try it for free first?

Yes. 2outube converts one video a day for free, so you can check the format on a single lecture before buying a pack. Plus ($4.99/month or $49/year) is the subscription option.

Does this include the code?

No. 2outube gives you the spoken text, not the code shown on screen; use Karpathy's repository for the code.