English as Code: Natural Language in the Era of Large Language Models

 

English as Code: Natural Language in the Era of Large Language Models

Inspirational foundation: Grant C. Sterling, Eastern Illinois University. Framing and analysis: Dave Kelly, 2026. Prose rendering: Claude (Anthropic).

---

The Thesis

In the era of large language models, English is the programming language. Precision in natural language is the discipline that was formerly called coding. Vague language is a bug. An instruction that is ambiguous, underspecified, or internally inconsistent does not merely communicate poorly — it executes poorly. The standard of correctness that once applied only to formal code now applies to prose.

This is not a metaphor. A natural language specification supplied to a large language model governs its outputs with the same logical force as a function call governs its return value. The governing document either specifies correctly or it does not. The LLM does not fill gaps from good intentions; it fills them from its training distribution, which is not the same thing. Every imprecision in a specification is an invitation to drift. Precision is the only defense.

---

The Demonstration

The Classical Stoic System built on Sterling’s philosophical corpus is the proof of concept. The system consists of a suite of analytical instruments — audit procedures, decision frameworks, evaluative protocols — written entirely in English and supplied to a large language model as governing specifications. The instruments produce consistent, structured, auditable outputs.

They do this not because the model was trained for philosophical analysis. They do it because the English specifications are written with sufficient precision to function as executable logic. The corpus is a codebase. The instrument definitions are functions. The self-audit requirements are assertion statements. The mandatory self-audit at each step transition is a checkpoint that halts execution and demands verification before proceeding. The prohibition on named failure modes is a guard clause. The five verdict categories of the Classical Ideological Audit are an enumerated type. The six philosophical commitments are constants against which variables are evaluated.

Sterling’s philosophical texts are the inspirational foundation of this system. They supplied the load-bearing concepts, the theoretical architecture, and the precision of argument that made instrument construction possible. A vague philosophical framework cannot be coded. Sterling’s framework is not vague. Its precision at the level of philosophical argument translates directly into precision at the level of executable specification. That translation is the work of instrument architecture.

---

The Implication

Any knowledge system intended to govern large language model output must be built to the standard of code. This means: terms defined before use, procedures specified in full, failure modes named and guarded against, output formats determined in advance, and self-audit steps written in at every transition point where drift can occur.

The humanist discipline of clear prose and the programmer’s discipline of executable specification have converged. The writer who wants his knowledge system to run correctly in the LLM era must think like an architect: structure before content, specification before execution, verification built into the procedure rather than applied after the fact.

This convergence changes what it means to write well. Clarity has always been a virtue of prose. In the LLM era it is also a technical requirement. The sentence that cannot be executed without ambiguity is not merely unclear — it is broken. The knowledge system that relies on the reader’s good judgment to fill in what it leaves unspecified has, in an LLM context, no reader in that sense. It has an executor. What the specification does not say, the executor will supply from elsewhere. Precision is the only guarantee that “elsewhere” is not where the output comes from.

---

Inspirational foundation: Grant C. Sterling, Eastern Illinois University. Framing and analysis: Dave Kelly, 2026. Prose rendering: Claude (Anthropic).