Hukam Is Not Command: Untranslatable Glossaries for AI

Every translation of a sacred term is a decision about what the term is permitted to mean. When a language model renders hukam as command, the decision has already been made — somewhere in training data nobody audited, at a scale no community consented to. This spine maps an essay on whether a tradition can keep jurisdiction over its own concepts inside machines built to flatten them.

I. The Problem Inside the Machine

The essay opens with the mechanism: large language models translate by statistical proximity, so the gloss that wins is the gloss that appeared most often, not the gloss that is right. Translation inside such systems is an ungoverned editorial act performed at industrial scale — and the traditions with the smallest digital footprints are edited the most.

II. Hukam Is Not Command

The worked example comes next, and it carries the essay. Command implies a commander external to the commanded — a sovereign issuing orders from above. The essay will argue that the Gurmat concept is closer to the operative order of reality itself, in the family of Asha and Rta, and that rendering it command quietly rewrites the metaphysics into an Abrahamic shape. One word, one flattened cosmology.

III. The Glossary as an Instrument of Sovereignty

The third section describes the proposed instrument: an untranslatable glossary, maintained by the community whose concepts are at stake. Each entry forbids the flattening gloss, states what the term does not mean, and marks the permissible semantic neighborhood. Conceptual sovereignty is defined here as definitional jurisdiction over a tradition’s own foundational vocabulary.

IV. Negative Constraint Prompts

The essay then turns technical. Negative constraint prompting and retrieval layers can forbid the model’s default gloss and supply the community’s constraints at the point of use. This section assesses honestly how far prompt-level intervention can reach before fine-tuning or corpus work is required — the essay promises an engineering audit, not a manifesto.

V. The Fellow Travelers

Precedent strengthens the case: jihad reduced to holy war, karma reduced to fate, dharma reduced to religion. Every tradition’s core term has been captured by its nearest Western neighbor. The glossary generalizes from a pattern already visible in the record rather than inventing a grievance.

VI. Governance, or the Gatekeeper Problem

The final section turns the Institute’s Masands Cycle on its own instrument. Every protective infrastructure becomes a gatekeeper unless watched: a glossary without accountable governance becomes one more committee editing transmission. The essay specifies who maintains the glossary, how disputes are arbitrated, and how the instrument is kept from hardening into the thing it resists.

Counter-case

The strongest opposing view: perfect translation is impossible, but useful translation is not, and glosses like command are entry ramps rather than final meanings — learners begin from the approximation and refine it. Hard refusal of translation isolates traditions and abandons the field to worse translators, and the community-consensus requirement is romantic: traditions disagree internally, so any glossary will encode one faction’s reading against another’s. On this view the model’s statistical average is no worse than the committee’s decree, and constraint prompts are censorship with better branding. The essay must show that its instrument survives the same gatekeeper critique it applies to the machine.

Sources to verify

• Arvind-Pal Mandair, Religion and the Specter of the West — UNVERIFIED, confirm before citing in the finished essay
• Barbara Cassin, ed., Dictionary of Untranslatables — UNVERIFIED, confirm before citing in the finished essay
• Anna Wierzbicka, work on semantic primes and cross-cultural semantics — UNVERIFIED, confirm before citing in the finished essay
• Emily Bender and collaborators, critiques of large language model meaning claims — UNVERIFIED, confirm before citing in the finished essay
• Standard lexicographic resources on hukam in Gurmat — UNVERIFIED, confirm before citing in the finished essay
• Technical literature on retrieval-augmented generation and constrained decoding — UNVERIFIED, confirm before citing in the finished essay

Stakes

This essay operationalizes the premise of the Institute’s Sacred Indic Critical Intelligence framework — that Indic categories function as analytic primitives, not translations of Western ones — at the site where the next generation will actually encounter those categories. It invokes the Masands Cycle as both application and warning, and it gives the Institute’s Gurmat work a concrete instrument: not another essay about erasure, but a mechanism against it.