Working paper 01 · Updated 18 November 2026

Continuity
across separation.

An operating-system proposal for continuity across devices and interrupted connections.

Abstract

Orynth OS is a research initiative exploring a new operating system for the intelligence age. The ambition is an extensible foundation for technologies, environments and ways of living that future builders will define. This paper examines one initial question: whether identity, state and authority can retain useful continuity as devices, operators, interfaces and network conditions change.

An initial research direction combines persistent, addressable objects; explicit capability grants; a bounded execution runtime; and protocols for exchanging state across independently operated nodes. Future builders would define the systems and experiences above those foundations. This working paper sets out a hypothesis and a programme of experiments.

01 / The problem: continuity across boundaries

Modern networked systems support extraordinary forms of collaboration and computation. In many cloud applications, however, a service still determines the authoritative copy of data, the account through which it is reached and the operations available over it. Local-first research documents the resulting tension between collaboration, ownership and continued access. [1]

As people, autonomous processes and devices participate in longer-lived environments, continuity becomes a system-level question. What should remain available when an operator changes? Which work may continue when a connection disappears? How should a capability retain its meaning when its implementation is replaced?

This paper explores a common operating contract across these boundaries as one part of the wider Orynth OS research agenda. Existing approaches provide substantial foundations. The work ahead is to test their composition, expose incompatible assumptions and identify which additional primitives are justified.

02 / A falsifiable hypothesis

A small, extensible operating model could allow independently built systems to preserve authorised state and identity across changes of host, interface and connection, while making the limits of that continuity explicit.

The first test would use two separately operated runtimes and independently authored capabilities. A third builder would add a previously unplanned extension through published contracts. Migration, interruption and revocation tests would establish exactly which guarantees survive.

The hypothesis would need revision if the demonstrations depend on hidden coordination, require the original operator to remain available, lose documented state during replacement, or permit effects beyond the stated grants.

03 / The proposed operating model

A node is a logical participant with identity, state and capabilities. A device may host several nodes; an environment may span several devices. People and communities would govern their nodes and grant authority to software or machines through explicit rules.

Identity
A verifiable identifier, with a defined recovery and key-rotation process. A display name is a label resolved to that identity.
State
Versioned objects carrying provenance, ownership, retention and a declared consistency policy.
Capability
A contract describing permitted operations, input and output types, effects, dependencies, limits and failure behaviour.
Runtime
Execution that isolates components, checks grants, accounts for resources and records recoverable progress.
Exchange
Messages and state transfers with explicit ordering, retry, duplication and expiry rules.

These are starting assumptions for this experiment. The wider operating-system agenda can revise them as evidence emerges. A first implementation could run on existing host operating systems. Kernel development would be evaluated against measured hardware, isolation and scheduling requirements.

04 / Identity beyond a website address

Content addressing separates the identity of material from a particular server location. IPNS adds signed records that let a public-key-based name point to updated content. These mechanisms provide a useful starting point for verifiable identity and resolution. [2]

An Orynth OS client could resolve a node identifier through a defined discovery mechanism and validate its records. Human-readable names could be optional labels. Such a design would still require bootstrap peers or another discovery path, reachable carriers, persistent storage and key management.

The proposal concerns an operating and addressing model that could run over several network transports. The public download site uses HTTPS and DNS. The Developer Preview resolves private .os names locally inside the client; the hosted gateway remains a separate single-service experiment. Network-wide naming, a native exchange protocol and independent discovery remain future work.

05 / Useful work during separation

Bundle Protocol v7 describes an overlay that can store, carry and forward data across intermittent or delayed links. This offers a concrete basis for studying eventual delivery under explicit expiry and resource constraints. [3]

For suitable data models, Automerge demonstrates independent local edits that synchronise later. Its replicated structures address convergence; application meaning and the validity of consequential operations still require their own rules. [4]

Orynth OS would therefore require each operation to declare its coordination needs. A scientific observation might be recorded locally. A shared exclusive reservation might require a coordinating authority. A disconnected node could show a result as provisional until the required evidence arrives.

The CAP literature establishes a relevant boundary: during a network partition, a distributed service cannot guarantee both linearizable consistency and a response to every request. The model must expose its chosen behaviour. [5]

The distinction also appears in offline value transfer. A signed message can be exchanged locally, while establishing an accepted transaction history is a separate agreement problem. Bitcoin’s original paper treats signatures and double-spend resolution separately. A new operating layer would still need explicit trust, coordination and settlement assumptions for such a use case. [8]

06 / Authority and independent composition

A process would receive bounded authority over named resources and operations. Enforcement would sit in the runtime, storage and device adapters. Revocation during disconnection needs an explicit exposure window: a remote node can enforce a new revocation once it learns of it, while short-lived grants can bound continued access.

Component systems such as the WebAssembly Component Model provide typed interfaces between independently compiled modules. Orynth OS could evaluate this approach for portable execution. Behavioural semantics, version migration, trust and resource limits would need additional contracts and tests. [6]

The platform test is independent invention: a new participant should be able to create a capability, replace an interface or introduce an object type through the published extension model. Documented compatibility would determine which existing systems can use it.

07 / Initial research programme

  1. Continuity. Run the same documented object on two hosts. Interrupt connectivity, restart both and compare recovered versions, provenance, lost updates and recovery time.
  2. Authority under delay. Expire or revoke grants while peers are separated. Measure the maximum period and effects of continued access. Verify resource limits and denied operations.
  3. Independent extension. Ask separate builders to implement compatible capabilities. Replace one and migrate its state. Measure contract deviations and the work required to integrate an unforeseen extension.
  4. Transport variation. Test connected, intermittent and delayed delivery with bounded storage. Record duplicates, stale state, expiry, eventual delivery and outcomes requiring reconciliation.

Reports should publish workloads, assumptions, failure cases and reproducible results. The comparison should include existing approaches that already solve parts of the problem. Architectural additions would be justified by measured gains and their costs.

08 / A frontier with physical constraints

Persistent research environments, collective systems and interfaces spanning physical devices are possible directions for independent experimentation. Long-delay communication offers a demanding boundary case: NASA’s DTN work provides an established lineage for networking across space-mission disruptions. [7]

Interplanetary scenarios could eventually stress-test continuity, autonomy and delayed authority. Distance, energy, storage, available links and the speed of light remain physical constraints. Any claim of suitability would require specialised validation against a defined environment.

The larger ambition is an extensible foundation whose future uses remain open. Its value would be established by what independent communities can build, sustain and govern on top of it.

09 / Status and stewardship

As of 18 November 2026, the repository contains a research interface, a hosted gateway experiment and an Orynth OS Developer Preview. The local client resolves installed .os addresses and runs app logic in isolated workers. Device identity and app data persist locally, with explicit capability approval and revocation. The gateway still contains three simulated logical nodes. Distributed execution, shared naming and cross-device synchronisation remain proposed work.

Research initiative describes the direction of the project. Legal form, governance, licensing and a sustainable funding model remain decisions to develop. This paper is an initial working proposal for scrutiny and experimentation.

References

  1. 01Local-first softwareKleppmann, Wiggins, van Hardenberg & McGranaghan · Onward! 2019
  2. 02Content addressing and IPNSIPFS documentation
  3. 03Bundle Protocol Version 7Burleigh, Fall & Birrane · IETF RFC 9171 · 2022
  4. 04AutomergeAutomerge project documentation
  5. 05Perspectives on the CAP TheoremSeth Gilbert & Nancy Lynch · 2012
  6. 06The WebAssembly Component ModelBytecode Alliance documentation
  7. 07Delay/Disruption Tolerant NetworkingNASA · Communications with missions
  8. 08Bitcoin: A Peer-to-Peer Electronic Cash SystemSatoshi Nakamoto · 2008