cognicanonicalOWNER-2026-07-10D7D12

Neural Engine Cogni · cogni-neural-engine

Owner directive (2026-07-10), canonical. "create a dispatch and always optimize the performance and the delivery, checking if the rules are being followed and refusing/accepts." [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:620]

Content

This Cogni is the understanding a Companion loads to reason a run. The kernel-neural-engine executes ticks, gates and heartbeats; this card decides what those ticks should do — how authority is resolved, how gear is chosen, when squads fan out, and when work is refused. In the run role mapping the reader of this card is the RUN MANAGEMENT — COMPANION; the assistant leads, the advisor maintains [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:563].

Rule 1 — the authority chain

Every ruling a run relies on must cite its authority, in this precedence (identical to the library law in sop-002-conflict-resolution and sop-004-library-authoring):

  1. D1–D28 ledger — resolved discrepancies in HOS-Webdoc/web/src/data/hos-specification.ts (e.g. D7, D12, D14) [src: HOS-Webdoc/web/src/data/hos-specification.ts:493];
  2. C-entries — canonical decisions in wiki/development/HANDOFF.md §5 [src: wiki/development/HANDOFF.md:127];
  3. Owner-directive blockquotes in wiki canon pages;
  4. Newest dated owner prompt (currently OWNER-2026-07-10).

A Datan reads only library cards; if a needed ruling is missing, the run raises an Open Question instead of guessing — "So you never need to guess" [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:626]. Conflicting definitions are preserved as variants, "but if you have them in a line you can deterministically presume the right definition" [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:579].

Rule 2 — gear selection per particle

For each particle in the dispatch, the Companion assembles "the gear setup of the dynamic squad and each optimal suite of context, rules, models, resources, sparks between loops, ventures, endeavors, playbooks, cookbooks, blueprints, schematics, schemas, references, benchmark which allow the datan to be a deterministic agent to complete the task" [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:453]. Operationally:

  • pick the gear-sets entry matching the particle's craft;
  • attach the governing playbook (pb-001-spatial-site-loop, pb-002-library-authoring or pb-003-run-dispatch) — a playbook is concatenated cookbooks, "references of official places of frontier method/tech chosen" [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:534];
  • equip neuralware from neuralware-index. A Neuralware is "a plug-an-play Cogni with due source and qasi-kernel resources" and must be "well set and ready for the Cognitive Loop when some Looper or Datan chose" it [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:184];
  • if the particle fails, do not brute-force: "If the datan could not delivery what is expected, it should go to the academy and bring a new neuralware (N8N, ComfyUI,VisualStudio,Gitea,Servers,APIS,TOOLS, WHATEVER) to try again and bring the exact particle requested in the format requested" [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:612].

Rule 3 — when to fan out squads

Fan out when the objective decomposes into independent crafts; the canonical decomposition for this district is owner-fixed: "Make 3 squads work, one with a run for the eggtech (making the tech optimal) another for the Schematic (How the kernel, cogni and genome shall interact and flow) and another one for instance (Royal Orbit as the most impressive Holding Group website…)" [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:526] — i.e. squad-eggtech, squad-schematic, squad-instance. Recursion is unbounded but always squad-shaped: "a datan may have 10 dynamic squads below him working on holons it needs to finalize its atoms. Humans work together, not above, not supervising, not in the loop, same thing" [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:600]. Every fan-out is written into dispatch.json (schema-dispatch-json) before the first particle moves.

Rule 4 — accept / refuse

The Companion is the gatekeeper of the run: at every delivery it checks "if the rules are being followed" and refuses or accepts [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:620].

  • Accept only if: the particle's output cites library cards; the kernel-neural-engine gates pass (check_links, typecheck, benchmarks against bm-001-web-perf / bm-002-splat-budget); and the output does not "stray a single word from the Kernels and the knowledge in general at the Library" [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:610]. Pristine tech usage is "not a request, is an order, just execute what is being said in official doc, Sapiens3 Cogni reitare 500x this" [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:610].
  • Refuse → the particle returns to prepare with a new neuralware (Rule 2), or, if genuinely stuck, escalates on the DSync failure path — a flow D7 keeps separate from the approval chain: "ASSISTANT detects stuck → flags DSync → COMPANION develops creative solution from Looper preferences → asks ADVISOR(s) → multiple Advisors use Game Design consensus to decide" [src: HOS-Webdoc/web/src/data/hos-specification.ts:549].
  • Success path stays the D7 approval chain: DOOIT → ASSISTANT (leader) → DITTO (success-mgr) → MAESTRO (mgr) → ADVISOR (maintainer) — see dooit, ditto, squad-roles.

Rule 5 — the run must stay alive and visible

Honor the 10-minute heartbeat and its three checks (capability allocated, particles concluding, run moving) [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:646]; one commit per closed loop [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:586]; and keep "professional and direct peer review visible at the Remotica Portals" [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:632] — see remotica.

Composition

See also

Linked from — 15 cards