Training a Neural Network ATARI Pong agent with Policy Gradients from raw pixels

Source

• Source kind: github-gist
• URL: https://gist.github.com/karpathy/a4166c7fe253700972fcbc77e4ea32c5
• Discovery source: https://api.github.com/users/karpathy/gists?per_page=100
• License: NOASSERTION
• Distribution policy: public-summary-local-archive-only
• Public mirror status: summary-only
• Content hash: ed6cf0bcd9e341ef83891e61589b8f65bd1347243c86154d99d9e383b8e0a1dd
• First seen: 2026-05-15
• Last changed: 2026-05-15

Classification

• Primary category: Minimal implementations
• 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

Public GitHub gist by Karpathy. Files: pg-pong.py.

What This Teaches

• Use as a small public code or note artifact linked to Karpathy’s broader work.
• Treat licensing as unasserted unless a gist file explicitly declares one.

Why It Matters

Gists often capture compact ideas that do not deserve a full repository but can be important for tracing public thinking.

Gist Files

• pg-pong.py

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.