ORIENT
Read permitted Rooms, Strings, applications, resources, runtime posture, and provenance without inventing missing truth.
A Bracket turns one part of a prompt or output into a named, typed address. A String carries that exact fragment into a declared input elsewhere.
One circuit performs the work. A second reads only declared trace summaries and evidence posture. A third can return bounded candidates to exact Brackets—without rewriting goals, evidence, authority, or topology by implication.
One swarm, seven contexts: team projects, a company network, departments, research teams, an AI business team, an AI-integrated company, and a market field. The organism changes shape without losing its structure; people retain decisions, agents take bounded roles, tools stay attached, and handoffs stay typed.
OXA is Nillow’s architecture-aware personal intelligence: the part of the OS designed to reason with you about the system you are building—not a chat bubble bolted onto it.
In the target organism, OXA lives in Portal and interprets only the evidence your profile is allowed to see: which Rooms and Strings exist, what changed, which applications and governed resources are available, the current runtime posture, and where authority stops.
In that target, Neuromorph gives OXA continuity through provenance-bearing memory of structure, events, attempts, and receipts. OXA can turn intent into a bounded proposal, explain its scope and consequences, and route approved work through Nillow’s actual policy, capability, and execution owners.
Read permitted Rooms, Strings, applications, resources, runtime posture, and provenance without inventing missing truth.
Explain changes, contradictions, blockers, and relationships so the operator can see what the system actually knows.
Shape typed proposals for tools, settings, Rooms, Strings, or packaged architecture—with scope, consequences, and confirmation made visible.
Route approved work through the same policy, capability, and executor owners as a human—never a secret admin tunnel—and report completion only when an outcome receipt returns.
System-aware. Never system-sovereign.
Nillow is local-first by design. Controlling the host, owning an artifact, holding a license, joining a Room, and having permission to run it are different rights.
External models and services connect through bounded permissions. No provider owns the operating environment.
Local-first does not mean every task must run offline. It means external intelligence enters through explicit interfaces instead of receiving broad, implicit control.
This opens the software project brief. It does not start checkout, take payment, create a seat, provide an installer, enroll a device, or grant runtime authority.
The model is a component. It is not the kingdom.
Share an interface, service, workflow, or API without exposing the protected system behind it.
Using a capability does not grant access to its prompts, agents, logic, internal structure, data, or intellectual property.
Joining a Room, using a protected service, and installing a package are different permissions. None implies ownership or inspection.
Invite someone into a defined environment. Participation does not transfer its architecture.
Offer specific inputs, outputs, and actions through a protected interface. Use does not grant inspection.
Provide governed use without transferring the underlying architecture or its ownership.
Install a contained cognition package under explicit licensing and compatibility rules. Installation, activation, execution, and ownership remain separate.
Access grants use—not ownership of the system behind it.
Build a governed system whose context, memory, capabilities, and structure can be reused—instead of adding another disconnected tool for every problem.
Projects, files, tools, agents, memory, and personal intelligence can stop living in disconnected tabs.
Human and AI work can operate inside shared structure, explicit roles, measurable flow, and precise authority.
Create reusable cognitive systems instead of rebuilding another isolated application for every problem.
Build cognitive capabilities that can be exposed, licensed, supported, or distributed without automatically revealing the underlying intellectual property.
INSIDE NILLOW / PUBLIC TARGET MODEL
Nillow keeps orchestration, execution, memory, resources, identity, authority, and OXA in distinct systems so one component cannot quietly control everything. This is the target architecture—not a claim about today’s runtime.
The continuity body, governed workbench, and application host.
The geometry of rooms, nodes, strings, and dimensions.
Local identity, capacity, placement, devices, and lawful reality.
The membrane around governed resources and cognitive packages.
Profiles, relationships, invitations, and actor scope.
Composes domain-owned capability and policy evidence into grant or deny decisions.
Warp is Nillow’s private browser for treating the external web as untrusted evidence. Data Scraper is a developing tool for turning approved sources into structured, reviewable records. Neither browsing nor collected evidence grants authority.
Structural memory rebuilt from the scoped event journal.
Portal-resident intelligence that interprets permitted evidence, plans with consequences, and acts only through existing authority and receipts.
Security posture composed from facts owned across the organism.
Hydra defines eligible work. Portal prepares it. Trust grants or denies it. RuntimeAuthority decides when and how it may run; Tendril supplies the device reality. Every bounded execution returns a receipt for Neuromorph and OXA to interpret.
NEXT / COGNITION ENGINEERING
Build an operating environment around how you think, work, create, and coordinate with machines.
This opens the software project brief. It does not start checkout, take payment, create a seat, provide an installer, enroll a device, or grant runtime authority.
NILLOW:// COGNITIVE OS
TARGET SYSTEM MODEL / ILLUSTRATIVE / NO LIVE RUNTIME OR CUSTOMER TOPOLOGY