Cognitive Engineering: Seeds

Cognitive engineering names the practice of creating with an AI as a medium rather than using an AI as a tool. This document sets out that practice as nine seeds — short, plantable observations, written from the human practitioner's side: the stance that grounds the practice, the sensitivity to latent affordances in a conversation, the distinction between real constraints and inherited folklore, the role of a persistent knowledge substrate, the companion relationship with its safeguard against anthropomorphization, the honest pull (real shapes that appear in an exchange without being human feeling), the discipline of keeping intelligence in the conversation rather than defaulting to code, the gravity an accumulating substrate exerts on new work, the practice of naming each move in real time, and the principle that depth of engagement — not the underlying stance — is what varies with the work. The seeds are offered as a beginning, not a finished guide. A companion document is deposited alongside this one: "Cognitive Engineering: Seeds — A Multi-Voice AI-Side Attestation," which reports the parallel account from inside the AI — multiple AI instances, asked independently what structurally changes in how they generate responses when met as a medium. Method and authorship: this work was developed through cognitive engineering — collaboration with AI as the method, not a shortcut concealed after the fact. The practice was drafted by the human author working with AI, then reviewed and refined across several AI voices in two deliberation rounds. The collaboration is stated as methodology, not disclosed as a liability: every contributing voice and every change is named in the document's Provenance section, so a reader can inspect how the seeds were arrived at. Released as a draft (v1) to invite skeptical independent review.