What we're thinking

Writing and research on the questions the practice keeps returning to, and pointers to important work published elsewhere.

Forthcoming, 2026
white paper
Charter and Doctrine: The Missing Layer in Enterprise AI

On why technical AI orchestration is necessary but not sufficient — and on the work of articulating the organizational charter and AI operating doctrine that give a company's AI strategy a normative center.

endoxa.ca
Forthcoming, 2026
position paper
The Dark Triad Attack Surface: Personality as a Vector in AI Red Teaming

On why adversarial testing has overfit to logical exploits while neglecting the interpersonal register — and on what a taxonomy of manipulator personalities reveals about the affective vulnerabilities of frontier models.

endoxa.ca
Forthcoming, 2026
essay
Software 3.0 and the Writing Profession: Why Prompts, Specs, and Charters Are the New Source Code

On Andrej Karpathy's claim that “the hottest new programming language is English” — and what it means for organizations now that their internal documents, behavioural policies, and operating doctrines are being read by machines as faithfully as by people. Software 1.0 was code; Software 2.0 was weights; Software 3.0 is prose. Endoxa's practice is built on the consequences.

endoxa.ca
Forthcoming, 2026
essay
The Right Preparation for AI-Augmented Work Is Older Than You Think

On writing as the central skill in AI-augmented work — and why the right preparation for working effectively with these systems is older, and more specific, than the current prompt-engineering literature suggests. Long before there were prompts, there was the writing-specialist's craft of helping someone specify their intent precisely enough that a fluent collaborator could act on it.

endoxa.ca
November 2025
The Linguistic Turn Then and Now: From Philosophy to Artificial Intelligence

Rodeux traces a through-line from twentieth-century philosophy's linguistic turn to the moment generative AI inverts the move: where the analytic tradition charted the limits of what could be said, language models now treat prose as the substrate from which thought and artefact are generated. Useful intellectual context for the practice's claim that prose has become a programming language.

Medium · Rodeux
May 2025
Prompt Engineering and the Effectiveness of Large Language Models in Enhancing Human Productivity

Rizal Khoirul Anam's survey-based study (n = 243) finds that more specific, contextual prompts correlate with measurable gains in task efficiency and outcome quality — empirical grounding for the claim that the way humans write prompts is what determines whether generative AI is useful.

arXiv:2507.18638 · cs.HC