Let’s reproduce GPT-2 (124M)
Source
- Source kind:
youtube-video - URL: https://www.youtube.com/watch?v=l8pRSuU81PU
- Discovery source: https://www.youtube.com/feeds/videos.xml?channel_id=UCXUPKJO5MZQN11PqgIvyuvQ
- License:
YouTube-standard - Distribution policy:
public-metadata-summary-only - Public mirror status:
summary-only - Content hash:
3714dad567184fdf461988c36398fe5fd273b9c982cd9d4878dd3667cfa05ca4 - First seen: 2026-05-15
- Last changed: 2026-05-15
Classification
- Primary category: Talks, courses, and videos
- Corpus source note: 2026-05-15-karpathy-public-corpus
- Project taxonomy: karpathy-project-taxonomy
- Idea map: karpathy-idea-map
- Topic hub: karpathy-public-work
Summary
YouTube video from Karpathy’s public channel: Let’s reproduce GPT-2 (124M).
What This Teaches
- Use as a lecture-style source for Karpathy’s from-scratch explanations and AI system intuition.
- Pair with linked GitHub repos when a lecture has an accompanying codebase.
Why It Matters
These videos are among the clearest public sources for Karpathy’s pedagogy and implementation taste.
Public Handling Notes
- EXTRACTED: This page records public metadata and a source-grounded summary.
- INFERRED: Full local preservation, when available, is for private/local use unless a license or explicit source policy makes public redistribution safe.
- Do not treat this page as permission to republish unlicensed source text or code wholesale.