Two different layers
In the architecture proposed in this brief, Gonka represents distributed AI compute and reduced dependence on centralized compute infrastructure. ObliNet represents observable Authority, Commitments, Dependencies and Accountability. The pairing remains a proposal that requires technical validation.
Gonka represents distributed AI compute in this proposed architecture. A separate ObliNet layer represents Authority, Commitments, Dependencies and Accountability. A dashed, non-directional line denotes a conceptual relation only. No existing integration or partnership is asserted.
In ObliNet, the commitment is the primary object of coordination. Its feasibility is checked before confirmation; afterward, dependencies, execution, changing conditions and confirmation of the result are tracked. AI helps participants formulate, check and align commitments, and assess evidence of their fulfillment. The authority to enter into or confirm commitments, and to accept fulfillment, remains with the relevant authorized participants.
Technological separation of powers
Decision Authority, AI Advisory Power, Identity, Data, Compute, Commitment Layer, Execution, Audit and Public / Stakeholder Oversight should have distinct responsibilities. No single actor should control all layers. Distribution of compute alone does not establish accountable governance.
No Single Irreplaceable Actor
What if a leader becomes corrupt, a model becomes unreliable, a compute provider blocks access, an auditor becomes dependent or ObliNet itself is captured? Leaders, models, providers, auditors and even the ObliNet implementation must be replaceable.
The protocol matters more than the platform
Portable history, independently inspectable records and replaceable implementations are design goals. Their practical value depends on interoperability, privacy, governance and credible exit procedures; they are not guaranteed by a diagram.
Why we propose working with the Liberman brothers
The rationale for approaching David and Daniil Liberman is to test the complementarity of Gonka’s compute infrastructure and ObliNet’s coordination model. Gonka describes a distributed AI compute network with developer access to inference through an API. ObliNet proposes organizing the use of AI around feasible commitments and verifiable outcomes.
An assistant to a public leader should not make one provider an irreplaceable source of advice. Distributed infrastructure could be part of this architecture, but does not itself guarantee model objectivity, data confidentiality or independent audit. Portability, verification rules and the ability to switch providers must be implemented and tested separately.
The proposed pilot models a multilevel city program to improve public transport accessibility: leadership sets goals and constraints, agencies coordinate budgets and infrastructure, operators and contractors accept interdependent commitments, and authorized reviewers assess outcomes for residents. A budget change or infrastructure delay should reveal consequences at every affected level and trigger renegotiation.
Using synthetic or open data, we propose comparing ordinary management, ordinary management with AI, and ObliNet with the same AI under comparable resource constraints. We would measure contradictions detected before commitments are accepted, adaptation time, outcome quality and the combined cost of errors and checks. Any advantage of ObliNet remains to be established experimentally.
For Gonka, this could provide an application demonstration and infrastructure requirements; for ObliNet, an opportunity to test the approach and implementation portability with a technical partner. This is a cooperation proposal: no agreed partnership, working integration or results from this pilot are claimed.