

AI Master's Popup: Neural Networks (building gpt tokenizer p2)
THIS IS A DAILY EVENT (weekdays)
New content every day.
This event is not beginner-friendly — background knowledge is mostly required (though experience levels vary). Feel free to message me if you're unsure. The goal is to cover top university master’s-level content in less time.
This is a group class, no official teacher. We currently follow Karpathy; pause, ask questions, and answer ourselves. Be respectful. The class is small.
We'll be discussing an ongoing series of AI-learning sessions, initially focused on Andrej Karpathy's Zero-to-Hero Neural Network playlist — covering GPT-2, backpropagation, tokenizers, and more. The course began partway through July. We'll finish the final two videos (GPT tokenizer + GPT-2 from scratch) in August, then shift to pre-training a GPT-2 model in class, followed by fine-tuning it into a chat assistant (see the end of Karpathy’s “GPT from scratch” video to understand the challenge of fine-tuning).
Playlist:
https://www.youtube.com/playlist?list=PLAqhIrjkxbuWI23v9cThsA9GvCAUhRvKZ
If you're new to NS in August (or were around but didn’t join), you’re still welcome! Ideally, watch the full 6-hour tokenizer + GPT-2 set at the start of August if you want to join the pretraining work later.
Month 1 Curriculum (spills into early August):
https://www.notion.so/shonixyz/Advanced-AI-Learnathon-Zero-to-Hero-231d8b25071280ec9a41f7003532826e?source=copy_link
The July series focused on doing backprop by hand and understanding fundamentals deeply. You’ll get more out of it each time you rewatch.
If you want to git gud at AI, come through.