Glyph GPT |

𓉡𓀬𓉢

Balancing weights of Anubis...

𓅱 𓃀 𓋴 𓊃 𓏏 𓈓 𓎺 𓏤 𓏤 𓋴 𓈙 𓏏 𓇅 𓎺 𓏤 𓏤 𓊃 𓏏 𓂝 𓎼 𓏏 𓎟 𓎺 𓏤 𓏤 𓇠 𓂝 𓎼 𓏏 𓈓 𓎺 𓏤 𓏤 𓐍 𓈖 𓆑 𓈓 𓎺 𓏤 𓏤 𓇋 𓈙 𓂧 𓈓 𓎺 𓏤 𓏤 𓋴 𓄡 𓏏 𓇅 𓎺 𓏤 𓏤 𓊃 𓏏 𓇠 𓂝 𓎼 𓏏 𓎺 𓏤 𓏤 𓇠 𓂝 𓎼 𓏏 𓎺 𓏤 𓏤 𓃀 𓅡 𓏏 𓇠 𓎺 𓏤 𓏤 𓈖 𓃀 𓋴 𓆭 𓂝 𓎺 𓏤 𓏤 𓏏 𓈖 𓃀 𓋴 𓈓 𓎺 𓏤 𓏤 𓎛 𓅱 𓂝 𓇹 𓈓 𓎺 𓏤 𓏤 𓐍 𓏏 𓎟 𓃀 𓇜 𓈖 𓏏 𓏤 𓏤 𓆶 𓂋 𓎟 𓏤 𓎛 𓈖 𓎡 𓏏 𓂠 𓎟 𓏤 𓏤 𓐜 𓋴 𓏑 𓎛 𓅱 𓏑 𓏑 𓏊 𓏠 𓏌 𓏏 𓅿 𓍿 𓂋 𓊪 𓅿 𓎟
2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 por 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 portions, 2 por 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
Text copied

Gardiner codes: G43 D58 S29 O34 X1 N33A W10 Z1 Z1 S29 N37 X1 M13 W10 Z1 Z1 O34 X1 D36 W11 X1 V30 W10 Z1 Z1 M33 D36 W11 X1 N33A W10 Z1 Z1 Aa1 N35 I9 N33A W10 Z1 Z1 M17 N37 D46 N33A W10 Z1 Z1 S29 F32 X1 M13 W10 Z1 Z1 O34 X1 M33 D36 W11 X1 W10 Z1 Z1 M33 D36 W11 X1 W10 Z1 Z1 D58 G29 X1 M33 W10 Z1 Z1 N35 D58 S29 M1 D36 W10 Z1 Z1 X1 N35 D58 S29 N33A W10 Z1 Z1 V28 G43 D36 N11 N33A W10 Z1 Z1 Aa1 X1 V30 D58 M30 N35 X1 Z1 Z1 M7 D21 V30 Z1 V28 N35 V31 X1 D39 V30 Z1 Z1 Aa14 S29 X3 V28 G43 X3 X3 W22 Y5 W24 X1 H1 V13 D21 Q3

How does it work?

GlyphGPT begins with nanoGPT, Andrej Karpathy's GPT architecture implementation, trained from scratch exclusively on hieroglyphic texts painstakingly prepared by Mattia De Cao. Optimized with Claude Code autoresearch. This generates streams of hieroglyphic sequences, which we then translate into English using Mattia's hiero-transformer - which iteself is based on Facebook's M2M-100 translation model, fine-tuned on the same hieroglyphics dataset. Essentially a mini AI learns to babble Middle and Old Egyptian, and big AI does its best to translate that into English. Both models have some room for improvement, but what could go wrong? Actually one thing for sure going wrong, Gardner notation has a larger vocabulary than unicode; so [Z3A] indicates the A variant of [Z3] but it is displayed as [𓏪A].

Contact

EgyptGPT GitHub

Thanks to Andrej Karpathy's for NanoGPT, and Mattia De Cao for releasing the translation model and dataset.