Can AI Beat a 1982 Commodore 64 Programmer?

Retro Hack Shack's latest upload examines whether modern AI can match a 1982 Commodore 64 programmer. The video tasks ChatGPT 5.6 Sol, Claude Fable 5 and Kimi K3 with writing a Space Invaders-style game using only BASIC V2. Each model's code is then evaluated against a human-written BASIC version from 1982.

The creator runs the AI-generated programs on an actual C64, records performance, and assigns scores based on functionality and fidelity to the original concept. The results are presented alongside the 1982 BASIC Space Invaders clone for direct comparison.

All BASIC source files are available in a linked GitHub repository, allowing viewers to review the code themselves. The video also includes brief product segments and timestamps for each segment of the test. Viewers can follow the progression from AI setup to final scoring within the video's chapter markers.

Watch the full AI versus 1982 programmer showdown now:

Comment highlights:

"Why not use the various APIs in Vice to provide access to the harness directly, instead of copy-pasting? I mainly use Hermes but this should work with OpenClaw also, if not all the AI apps. If it w..." - @agentmith