Andrej Karpathy's Deep Dive into LLMs: A Complete Walkthrough of How ChatGPT Works

Grounded / Real Inflated / Uruttu

95% real

5% uruttu

article Original Content

Dive into LLMs like ChatGPT (by Andrej Karpathy) This video tutorial covers - pretraining data (internet) - tokenization - neural network I/O - neural network internals - inference - training and inference - Llama 3.1 base model inference - pretraining to post-training - post-training data (conversations) - hallucinations, tool use, knowledge/working memory - knowledge of self - models need tokens to think - tokenization revisited: models struggle with spelling - jagged intelligence - supervised finetuning to reinforcement learning - reinforcement learning - DeepSeek-R1 - AlphaGo - reinforcement learning from human feedback (RLHF) - keeping track of LLMs - where to find LLMs Video lecture - https://lnkd.in/gV34nrXW

verified Validated Content

This checks out as an accurate description of a real, well-known video.

Confirmed accurate:

The video is real — "Deep Dive into LLMs like ChatGPT" by Andrej Karpathy, released February 2025, running about 3 hours 31 minutes, confirmed directly via Karpathy's own X/Twitter announcement and his YouTube channel.
Karpathy's credentials/context — he's a well-known AI researcher (co-founder of OpenAI, former Tesla senior director of AI), and this video is widely cited as one of the most substantive public explainers of how LLMs work.
The topic list closely matches the video's actual content: multiple independent summaries and note-takers (Medium, DEV Community, Substack writeups) confirm the video covers, in roughly this order: pretraining on internet data, tokenization, neural network internals and I/O, inference, the shift from pretraining to post-training, post-training on conversation data, hallucinations, tool use, the "models need tokens to think" concept, tokenization-related struggles like spelling, the supervised finetuning → reinforcement learning progression, RL examples like DeepSeek-R1 and AlphaGo, RLHF, and guidance on tracking/finding LLMs.
The specific listed items — "jagged intelligence," "knowledge of self," "Llama 3.1 base model inference" — are recognizable, specific concepts/segments that multiple independent viewers note as memorable parts of the actual video, not generic filler.

Verdict: This is simply a real, accurate table of contents for a real, widely-covered educational video. Nothing here appears fabricated or exaggerated — it reads as a faithful outline rather than a promotional oversell.