VerifyStack
← Back to Registry
85/100Verified
YouTube·Tutorial·

Let's build GPT: from scratch, in code, spelled out.

by Andrej Karpathy
View original on YouTube

Summary

This video explains ChatGPT's foundation as a probabilistic language model built on the Transformer architecture, introduced in the "Attention is All You Need" paper. The speaker outlines the process of building a simplified character-level Transformer language model from scratch, using the Tiny Shakespeare dataset, covering data preparation, tokenization, and the `nanogpt` codebase. The goal is to demystify how systems like ChatGPT function at a fundamental level.

Score Breakdown

Raw score: 85= 85/100

Automated Verification

40 / 40
Prompt Test10
Code Execution
Link Validation
Tool Claims Check8
Version Accuracy

AI Quality Analysis

36 / 40
Originality6
Specificity8
Completeness7
Value Density8
Honesty Limitations7
Model: anthropic/claude-sonnet-4

Context Signals

9 / 20
Freshness2
Author Track Record0
Genuine Engagement7

Verification Tests

PASSPrompt Testing1558ms
PASSPrompt Testing4380ms
PASSTool Claims Check12722ms