The executive core.
Admission, budgets, reasoning, synthesis and evidence review. Specialist regions advise; the executive core arbitrates.
Intended architectural reach
Multi-model reasoning.
Across machines.
Within boundaries.
Distributed Inference Cognition™ is the technology direction behind this vision: one governed system of reasoning across major model families and up to 250 authorized constellation nodes—from old laptops on a LAN to private infrastructure and FRONTIER cloud compute.
Architectural targets, not a list of certified integrations or provider partnerships.
THE INTRANET OF CELESTIAL NODES
The real stars
are participants.
Models, machines, solvers, data, tools and human expertise. Each admitted participant contributes a bounded capability to one governed mission.
NORTHSTARExecutive
corearbitration + synthesis
corearbitration + synthesis
ADMITTED CAPABILITY / 01
Models
Formulate, critique and synthesize complementary lines of inquiry.
- Identity
- Who is contributing?
- Scope
- What are they authorized to do?
- Budget
- What resources may they use?
- Evidence
- What must return with the result?
Connection is not consensus.
Capacity does not grant authority.
INFRASTRUCTURE WITH EXPLICIT BOUNDARIES
One mission.
Distinct responsibilities.
The intended architecture separates executive reasoning, computation, evidence and network control. Each boundary can be inspected and qualified independently.
EXECUTIVE COREEXTERNAL BOUNDARYSCOPED CAPABILITY
ARCHITECTURE ILLUSTRATION · OPTIONAL NETWORK FABRIC DISABLEDNetwork Fabric.
An optional lane, disabled by default. Network control and its effective state belong to the external system and require independent proof.
Nodes and executors.
Scoped models, compute, tools and approved infrastructure. Participation, authority and concurrency remain separate questions.
NORTHSTAR does not mutate network nodes, sign network commands, alter network evidence or determine effective network state.
02 / THE COMPUTE CONTINUUM
Different machines. Shared purpose.
From a local machine to private infrastructure and frontier compute, contributions are scoped by capability, policy, and the mission.
HUMAN PURPOSE
DISTRIBUTED INQUIRY
EVIDENCE & SYNTHESISNORTHSTAR / COMPUTECONCEPT VISUALIZATION
LOCAL MACHINESPRIVATE CAPACITYFRONTIER COMPUTESCOPED CONTRIBUTIONS
Roles before resources
How multi-model AI orchestration is intended to work.
A general-purpose model may investigate a question, a specialist may test a narrow claim, and a separate contributor may challenge the evidence. NORTHSTAR’s intended orchestration assigns these roles according to capability, independence, data permissions, budgets, and deadlines. The executive core remains responsible for arbitration and final synthesis.
Adding machines expands possible contributions; it does not increase their authority. Each model adapter, worker, tool, or external product needs a defined contract and environment-specific qualification. The 250-node proposal distinguishes logical participation from active concurrency and tests the coordination boundaries under failure.
Model breadth
No single model
defines the system.
General-purpose models, specialist models, open-weight models, commercial APIs, and private institutional endpoints—working together, and evaluating one another’s work.
Selection should consider demonstrated capability, evidence needs, available context, privacy, cost, deadlines, and independence. A model can investigate a subproblem, critique a finding, or contribute to synthesis without becoming the final authority.
Tools and MCP services belong behind their own admission and execution controls. Permission to read evidence is not permission to take an external action.
Elastic in more than one direction
A continuum of compute.
Not a fixed cluster.
“Omnidirectional” describes the intended ability to scale across resource count, capability, location, scheduling windows, and trust domains while keeping authority, provenance, budgets, recovery, and final publication under control.
Local and distributed
From a stack of old laptops on a LAN to private infrastructure and FRONTIER cloud compute: each capable, authorized machine has a place in the intended Distributed Inference Cognition™ fabric. Different nodes take on different work, rather than being treated as interchangeable.
Private and institutional
Enterprise and research clusters, on-premises infrastructure, private schedulers, reserved GPU queues, and air-gapped environments where separately supported and qualified.
FRONTIER cloud compute
The target includes major public cloud providers: queued GPU time, elastic instances, managed inference, and approved free-tier or paid-tier resources. Quota, billing, region, availability, and mission authority must be observed—not assumed.
Capability, not hardware prestige
Different hardware.
Useful work. One fabric.
The ambition is not to turn every laptop into a supercomputer. It is to make each capable, authorized machine a useful contributor to a larger inquiry.
Older laptops. Real roles.
A CPU-only node might run a compact model, retrieve and prepare evidence, or carry out bounded checks. Its role follows demonstrated capability, available memory, and the mission’s deadline.
More power. Heavier work.
A workstation or high-end GPU node might handle larger-model inference and compute-intensive specialist tasks. Its findings still need evaluation; more compute does not make an answer automatically correct.
Queued capacity. Shared purpose.
Private GPU schedulers and approved cloud allocations could join while their capacity is available. Budget, residency, trust, and admission remain explicit as the fabric expands or contracts.
Illustrative node roles, not universal hardware compatibility or a tested device list. A node, a model, an API endpoint, and a simultaneous inference call are different counts. Coordinating their contributions does not automatically pool their memory into one larger model.
Initial-release constellation target
1 Commander.
Up to 249 workers.
NORTHSTAR is being engineered to support constellations of up to 250 participating nodes while preserving a single accountable executive boundary.
1Commander
Coordinate—not outvote.
The Commander retains mission authority, budgets, evidence admission, recovery decisions, and final publication. Worker agreement alone does not become truth.
EXECUTIVE CONTROL249workers, maximum
Capability-matched contribution.
Workers may be models, machines, specialist systems, or approved services. They investigate, challenge, verify, and return evidence according to their actual capabilities and mission scope.
DISTRIBUTED CONTRIBUTIONBeyond 250future horizon
Grow only with proof.
Thousands remain a longer-term horizon as control-plane hardware, scheduling, evidence handling, recovery, and coordination are demonstrated at progressively larger scales.
FUTURE QUALIFICATIONNot 250 votes. Not a majority-rule chatbot. The 250-node target is an initial-release capability under engineering and final qualification—not a measured production-fleet claim.
Topology-aware by design
Not just the same LAN.
Not just the same cloud.
The optional Network Fabric lane is disabled by default. Separate, versioned product boundaries may expose approved evidence about private networking and reachability.
The system should reason from observed identity, reachability, topology, and policy, rather than assume a flat network. It must remain disconnected when no approved safe path exists.
Network planning and network enforcement are distinct responsibilities. The standalone Mainely Code network platform must establish its own simulation, staged enforcement, proof, and rollback before later product adoption. Inference workers do not inherit unrestricted network authority.
Windows-first and subsequent cross-platform implementations require separate backend and real-network qualification. An abstract contract or a website animation cannot establish packet forwarding, effective filtering, or tunnel interoperability.
Sovereign deployment direction
Control the boundary.
Not just the region.
Data
Decide what stays local, private, jurisdiction-bound, or eligible for external processing.
Identity
Bind nodes, services, approvals, and execution to verified identities.
Policy
Enforce tenant, budget, residency, and tool constraints before dispatch and publication.
Evidence
Retain an inspectable account of contribution, verification, uncertainty, and recovery.
Sovereign control is a design objective. Compliance, isolation, jurisdictional suitability, and security require deployment-specific assessment and evidence.
A term, not a product tier
What we mean by
FRONTIER cloud compute.
High-capacity cloud accelerators, managed inference, frontier-model endpoints, and queued compute admitted under explicit mission authority.
The term describes the upper end of the resource continuum. It does not imply a special commercial edition, unlimited access, provider endorsement, universal compatibility, or permission to bypass budgets, locality, identity, or evidence requirements.
Omnidirectional scaling
NORTHSTAR is designed to scale across resource count, capability, location, scheduling windows, and trust domains—not simply “outward” to more machines.
Follow the work
A greater horizon.
A shared conversation.
AI researchers, institutions, infrastructure teams, and people who care about where this science is going: follow the work and connect with Mainely Code.
Clear answers
Questions about the architecture
Does the website certify support for every model or cloud provider?
No. Model families and compute environments describe the intended architectural reach. Actual interoperability depends on the adapter, endpoint, credentials, configuration, and evidence for the version and environment being qualified.
Can a local machine and a cloud model contribute to one inquiry?
That is an architectural objective. Each contribution would need an admitted identity, a suitable capability, approved data access, and bounded budgets and deadlines. Protected work must remain inside the mission’s permitted locations.
Does connecting a tool give it permission to take actions?
No. Access to evidence and permission to execute an external effect are different boundaries. Product systems remain external arms with explicit, versioned contracts and separate authority.
Mainely Code’s co-flagships
Reason deeply. Build deliberately.
NORTHSTAR explores coordinated synthetic reasoning. Buildroom ADVANCED focuses on ambitious engineering workflows. Both are Mainely Code platforms in development, with distinct responsibilities.
A research conclusion might inform an engineering task, but it is not permission to execute one. Any future handoff must preserve an explicit, versioned boundary and separate qualification. Product systems remain external arms; NORTHSTAR’s executive core arbitrates while specialist regions advise.