The Flood and the Vessel: Semantic Preservation Across the Emoji Bridge

The Flood and the Vessel is a 2026 field study of lyric structure under emoji compression. Rebekah Cranes and Nobel Glas use Jack Feist’s poem “ARK” as the source object and ask what survives when the poem passes through a pictographic channel with little or no ordinary linguistic syntax.

Five AI systems independently translated the poem into emoji using different policies: annotated parallel text, dense technical glossing, pure glyph stream, architectural checksum, and liturgical repetition. A separate unprimed system then translated the emoji artifacts back into English without access to the source. The study’s principal reported finding is that every reconstruction recovered the same five-part motion: exhaustion → saturation → compression → transfiguration → ingathering.

Convergence among the encoders identifies the rose, spiral, butterfly or moth substitute, giraffe, salt, ark, light-and-zero pattern, and anatomical heart as structural anchors. The experiment also documents systematic loss. The butterfly beautifies the source poem’s moth-eaten decay; proper names, location, date, and Ginsberg lineage largely disappear; image survives more readily than connotation; and the poem remains more recoverable than its poet.

The round trip is also generative. Back-translations produce lines absent from the source, and the final synthetic poem incorporates new material when it recurs across witnesses. Cranes and Glas therefore distinguish generative compression from simple copying: the reconstructed poems are descendants constrained by the source architecture, not replacements for the original.

Within the archive, this study grounds later design claims about glyphic checksums and the Authoritative Emoji Map. It should be cited as a documented experiment on “ARK,” not as proof that all lyric poetry survives all lossy channels or that the models’ outputs were free of interpretive bias.