KAHRELUM OS

Architecture and doctrine

Evidence-to-decision control plane.

KAHRELUM OS is an application-layer operating system for research, AI evaluation, and decision work that must remain traceable and human-approved before action. It is not a kernel OS and not a chatbot product.

Separate facts from inference. State what remains unproven. Document constraints and falsifiers. Do not confuse model confidence with accuracy.

Core loop

  1. Intake — signals, documents, model outputs, OSINT, tickets.
  2. Classify — fact vs inference vs forecast vs opinion; source class.
  3. Verify — RECORD LOCK path: support, counterargument, revision conditions.
  4. Stress — optional red-team pass on high-impact claims.
  5. Gate — human approval for publication, deployment, outbound contact, spend.
  6. Act / record — action state and audit trail.

Intelligence pipeline

SCOUT (find)
  → BLACKGLASS (attack)
  → independent verify
  → KAHRELUM decide

SCOUT finds. BLACKGLASS attacks. A human gates publish, spend, and outreach. These are control-plane behaviors, not public SaaS products.

Modules

ModuleRole
RECORD LOCKEvidence verification and controlled publication
Intel TripwireStaged operating picture and source health
AI Red-Team DashboardAuthorized model-behavior testing
PortfolioPublic publication surface

Autonomy boundary

May run without extra approval: research, analysis, drafting, scoring, internal records, branch work, diagnostics, preparation.

Requires human approval: public publishing, job applications, recruiter outreach, production deploy, protected-branch merge, purchases, contracts, credentials, destructive actions, financial transfers.