IDKMesh explores how humans, AI agents, software tools, and distributed compute can collectively discover, design, build, verify, and operate useful systems even when the original objective is ambiguous.
The project begins from uncertainty rather than pretending the final product is already known.
Create a collective-intelligence and distributed-computing platform that can conceptually scale from one laptop to very large numbers of heterogeneous participating computers.
The system should support a loop such as:
goal -> hypotheses -> architectures -> tasks -> experiments -> implementations -> verification -> deployment -> learning -> improved goals
Humans and AI agents generate, criticize, test, compare, and select ideas.
Goals, questions, hypotheses, code, experiments, evidence, tests, dependencies, decisions, and provenance are represented as a graph.
Heterogeneous laptops, GPUs, servers, clusters, cloud resources, and edge devices execute work under different capabilities, latencies, trust levels, costs, and availability.
The project is not based on the assumption that more agents automatically produce better results.
A more useful working model is:
collective value = f(competence, diversity, independence, verification, specialization, coordination)
Many agents with correlated errors can confidently reproduce the same mistake. Therefore IDKMesh should explicitly measure and reward useful diversity and independent verification.
IDKMesh should investigate mechanisms that improve:
Can we design mathematical and software rules under which large numbers of imperfect humans, AI agents, and computers produce intelligence and software whose quality systematically improves as participation increases?
IDKMesh is not yet committed to:
The project should experimentally compare alternatives before making such commitments.