Sovereign Intelligence Systems
18 of 18
Chapter 18 of 18

Translation Notes & Sources

4 min read · 849 words

import { Footnote, ScholarNote, EndNote } from '@repo/sovereign-edition/components'

## On the Re-rendering of an English Translation

This edition occupies an unusual position in the history of Gilgamesh translation. Every English rendering of the epic since George Smith's first decipherment in 1872 has been a human scholar's work — sometimes drawing upon the work of predecessors, but always filtered through a single mind's judgment about what Akkadian means and how English should sound.

Our edition is different. We began from public domain English translations — principally Muss-Arnolt (1901), Jastrow & Clay (1920), and Kovacs (1989) — which themselves represent the best scholarship of their eras. We then subjected each tablet to the Seventy-One Council-and-Judge process, producing what might be called a re-rendering: a text that has passed through multiple literary sensibilities, been evaluated against multiple criteria, and converged on a version that balances fidelity with beauty.

This is not a new translation from the Akkadian. It could not be. The council does not read cuneiform. But it reads English with extraordinary sensitivity, and it can perceive — better than any single human reader — the places where a century-old translation's language has grown distant, where its rhythms have stiffened, where its imagery has lost its power to move.

The Akkadian text of Gilgamesh is riddled with gaps — places where the clay is broken, the signs effaced, the meaning lost. In the Standard Babylonian version compiled by George (2003), approximately one-third of the poem is reconstructed or missing entirely.

In this re-rendering, we have preserved the scholarly conventions of the source translations:

  • Square brackets [ ] enclose words that are restored from parallel passages, scholarly conjecture, or other manuscripts
  • Ellipsis marks … indicate unrecoverable gaps
  • Parenthetical question marks (?) mark uncertain readings

We have not attempted to fill gaps that the source translations left open. The poem survived three millennia in fragments; we honor that condition rather than smoothing it over.

The epic uses a syncretic pantheon of Sumerian and Akkadian gods. We have preserved the name forms used in the source translations rather than imposing a consistent transliteration scheme, as the scholarly sources from different periods used different conventions. The principal gods in the epic are:

  • Shamash (Sumerian: Utu) — Sun god, protector of Gilgamesh
  • Ishtar (Sumerian: Inanna) — Goddess of love and war, who proposes marriage to Gilgamesh and sends the Bull of Heaven when refused
  • Enlil (Sumerian: same) — Lord of the air, king of the gods
  • Ea (Sumerian: Enki) — God of wisdom and sweet waters, who warns Utnapishtim of the Flood
  • Anu (Sumerian: An) — Sky father, the most remote of the great gods
  • Bel — "Lord," an epithet of Enlil used throughout the epic
  • Nergal — God of the netherworld and pestilence
  • Ereshkigal — Queen of the netherworld, sister of Inanna/Ishtar

The complete evolution history for each tablet — including every proposed translation from every council seat, every judge's score, and the progression of convergence — is archived in the Seventy-One experiment corpus at:

sovereign-corpus/productions/seventy-one/gilgamesh/en/

Each experiment directory contains:

  • manifest.json — the translation parameters
  • steps/best/step_N/ — the best candidate at each evolutionary step
  • logs/evolve.log — the complete council-and-judge dialogue
  • pipeline_state.json — the pipeline's internal state at completion

These logs are the "scholarly apparatus" of this edition — the evidence that this translation was produced by a rigorous process rather than a single generation from a single model.

## Primary Sources

  • Muss-Arnolt, William. "The Gilgamesh Narrative, Usually Called the Babylonian Nimrod Epic." In Assyrian and Babylonian Literature: Selected Translations, edited by Robert Francis Harper, 324–368. New York: D. Appleton and Co., 1901.
  • Jastrow, Morris, Jr., and Albert T. Clay. An Old Babylonian Version of the Gilgamesh Epic on the Basis of Recently Discovered Texts. Yale Oriental Series, Researches IV, no. 3. New Haven: Yale University Press, 1920.
  • Kovacs, Maureen Gallery. The Epic of Gilgamesh. Translated with an introduction. Stanford: Stanford University Press, 1989. Electronic edition by Wolf Carnahan, 1998.

## Scholarly References

  • George, Andrew R. The Babylonian Gilgamesh Epic: Introduction, Critical Edition and Cuneiform Texts. 2 vols. Oxford: Oxford University Press, 2003.
  • George, Andrew R. The Epic of Gilgamesh: A New Translation. London: Penguin Books, 1999. (Revised edition published 2003.)
  • George, Andrew R. "SB Gilgamesh I, Score Transliteration." SOAS Research Repository, 2007. Available at soas.ac.uk/gilgamesh.
  • George, Andrew R. "The Gilgamesh Epic at Ugarit." Aula Orientalis 25 (2007): 237–254.
  • George, Andrew R. "Back to the Cedar Forest: The Beginning and End of Tablet V of the Standard Babylonian Epic of Gilgameš." Journal of Cuneiform Studies 66 (2014): 69–90.
  • Electronic Babylonian Library (eBL). LMU Munich. "Poem of Gilgameš: Text Edition." ebl.lmu.de/corpus/L/1/4.
  • Cuneiform Digital Library Initiative (CDLI). University of California, Los Angeles. cdli.ucla.edu.

## Technological Infrastructure

  • DeepSeek V4 Pro — Reasoning and literary judgment model. DeepSeek AI.
  • Mimo V2.5 Pro — Creative and literary generation model. Minimax.
  • LiteLLM Proxy — Model orchestration and routing. BerriAI.
  • BGE-M3 — Bilingual embedding model for retrieval. BAAI.
  • ASI-Evolve Engine — Council-and-Judge pipeline for iterative translation evolution. Consilience.
  • The Seventy-One Press — Translation service, manifest-driven. Consilience.

This edition was typeset in the Sovereign document model and published to the Consilience Press digital archive.

That is the whole work.

Take a copy with you, or go back to the contents and read a chapter again. Nothing here is gated, and nothing asks for a sign-up.