Coordinated Work: a research programme for organizational decisions
Published
Author: Talomnia team
Knowledge Contract opens Coordinated Work: needs, work networks and competencies linked to checkable decisions. 23 epics, 104 packages; announcements first.

Illustration supplied by Pavel Valentov for this announcement under recorded authorization. It depicts the programme's concepts; it is not an experimental result or scientific evidence. Image permission.
A team can finish more tasks while the organization delivers less. A department can raise its own output and increase the queue waiting for the next department. An agent asked to improve a local metric needs enough context to recognize those consequences before recommending a change.
Knowledge Contract is opening Coordinated Work, a research programme on how to represent an organization’s needs, work networks, resources, competencies and decision evidence. The proposed organizational profile connects this knowledge to the existing Agent Situation and Context Assembly work. Its contribution still has to be established against prior research and tested.
Source: Coordinated Work programme and intended contribution.
Why Talomnia needs this research
Programme author Pavel Valentov (X post) describes the starting point as follows:
For agent-based work in Talomnia, I needed to connect company needs, people's competencies, processes and resource allocation. There are many individual classifications and models. But I have not yet found an integrated system that meets our requirements.
For Talomnia, this raises questions about which knowledge and competencies a particular job actually requires, how people, agents and available resources fit together, and when there is insufficient information to make a recommendation. Talomnia's practical questions and the engineering work of the Arcanada ecosystem help formulate the programme's tasks and constraints. The proposed contribution still needs comparison with prior research, including coordination theory, and testing in practical cases.
What the programme contains
The proposal connects 23 epics, 104 work packages and 247 blocking dependencies. It preserves the original 72 steps and adds 32 explicit packages for gaps in the research and delivery plan. Each package records an output, acceptance criteria, a negative control and the predecessor outputs it consumes.
The scientific map identifies 16 potential papers and 12 hypotheses. Topics include needs and capabilities, resource semantics, work and flow networks, team composition, professional development, forecasts, knowledge sufficiency, justified decisions, causal evaluation, stable control, transfer and knowledge revision. These are proposed research boundaries, not completed publications or established results.
Source: Epic register, paper and hypothesis maps; machine-readable dependency graph.
How the work can proceed
First the programme appears in the Journal, roadmap and repositories; then source review, ontology and study design proceed in parallel.
Source review, ontology, provenance, data contracts and study design can then advance in parallel where their inputs permit. Manufacturing, service and project benchmark families follow their own prerequisites. Independent oracles and model-validity checks come before confirmatory comparisons. Planned comparisons include a strong operations-research approach, an agent without retrieval, ordinary RAG and the full Knowledge Contract profile, with declared budgets, ablations, uncertainty and full costs.
Each scientific release must meet the evidence obligations of its own claims. A literature review or conditional proof has a different release path from a claim about an intervention’s causal effect on an organization. Real-world use requires an eligible task, shadow evidence and a separately authorized bounded evaluation.
Source: Launch epic, parallel sequence and claim-specific release gates.
What evidence exists now
The starting register contains 1,176 bibliographic records: 157 checked primary records and 1,019 citation candidates. It does not establish 1,176 unique works or completed readings. Publication identities still need reconciliation, and completed full-text critical appraisals are zero.
The retained numerical probe covers 10 deterministic scenarios, 100 case-period observations and 190 elementary checks. It supports a bounded implementation claim under stated assumptions. It does not establish agent effectiveness, organizational validity or a successful field intervention.
Source: Starting evidence and limits.
Read and contribute
The programme proposal (a source proposal under review; its Contract Place page is forthcoming) brings together the full work inventory, paper boundaries, hypotheses and pinned source graph. The Journal announcement (forthcoming) will explain the organizational question in more detail. The Arcanada blog announcement carries the same three-site release. Materials are prepared in English, Russian and Simplified Chinese on Contract Place; this Talomnia post is English and Russian.
The existing programme proposal is the review entry point. Useful first contributions are the closest prior work, full-text appraisals, counterexamples, clearer resource and competency models, stronger experimental designs and independent reproduction. The result we want is a traceable path from a research question to evidence and a justified decision.
Sources
- https://github.com/Arcanada-one/knowledge-contract/blob/1e9aba657a580535cea4487c87fede13dff28c92/roadmap/coordinated-work/plan/PROGRAMME.md
- https://github.com/Arcanada-one/knowledge-contract/blob/1e9aba657a580535cea4487c87fede13dff28c92/roadmap/coordinated-work/plan/execution-graph.json
- https://arcanada.ai/en/blog/coordinated-work-2026-11-18
- https://github.com/Arcanada-one/knowledge-contract/pull/125
- https://x.com/VeritasArcanaAI/status/2123093943373955265