kalina.labs
Method / research-operating-systemDocumented state

Research Operating System

A versionable system for AI-assisted research.

ROS makes the research process traceable: research questions and collection lead to a documented substrate from which analysis, hypotheses and iterations emerge. ROS 1.7.2 adds context-dependent process diagrams bound to the source state to the consolidated Research State, with separate semantic and artifact checks.

Version
1.7.2
Updated
2026-10
Output
Method specification

Process model

Information chain

  1. 01

    Premises & Research Brief

    Define ResearchQuestion, terminology, system boundary, initial assumptions and ScopeProfile. COMPACT and FULL pursue different research objectives.

  2. 02

    Collection & Source Population

    Record CollectionScope, source origin, perspective, observations, measurements and explicit gaps; distinguish echoes from independent evidence.

  3. 03

    Substrate & Context

    SubstrateAssembly makes the documented source material traceable. Material, technical, energy and economic constraints, temporal reference and knowledge_time are assessed before hypotheses are formulated.

  4. 04

    Analysis, Systems Lens & Graph Resolution

    Examine relationships, mechanisms and dependencies. Following a positive Applicability Gate, the Dynamic Systems & Network Lens formalizes topology, states, flows, feedback, delays and time scales at levels ranging from implicit to formal/numerical. The Domain-System Graph and epistemic Research Graph remain separate.

  5. 05

    Hypotheses & Adversarial Review

    In the FULL path, test explanations against countermodels, falsifiers and AdversarialAssessment. Internal Predictions and ExternalForecast remain distinct; COMPACT does not require hypotheses.

  6. 06

    Signals, Iteration & Output

    Where the signal/index module applies, track temporal patterns with RAW/Derived separation and DerivationEdge. New evidence enters ResearchIteration; approved outputs remain separate from the canonical state.

Quality gates

Rules for robust deductions

  • Keep primary and secondary sources separate.
  • Label facts, inferences, hypotheses and normative evaluations.
  • Do not smooth over contradictions; preserve them as research objects.
  • Publish each revision with open questions and review status.
  • ScopeProfile determines which modules and checks apply; COMPACT does not require complete hypothesis or graph analysis.
  • knowledge_time and source independence prevent hindsight bias and double-counting of echoed evidence.
  • The Dynamic Systems & Network Lens is used only after an Applicability Gate and only at the level needed to produce insight.
  • The Domain-System Graph, process model and epistemic Research Graph must not be presented as a single unified graph.
  • Check missing material drivers, HARD/SOFT/UNKNOWN constraints and countervailing effects; do not present description as an identified causal effect.
  • Keep append-only history distinct from the consolidated current state; generate complete registers from the same fingerprint and subject red-team findings to scrutiny as well.
  • Assess material process branches for flowchart applicability; do not equate process transitions, physical causality and evidence. Review diagram, text alternative and export separately.

Operational components

Related system components

Coupled systems and domain modules with a shared identity and provenance model.

Knowledge SystemsActive

Nea Polis BIS

Coupled business-intelligence component for organization-specific facts, metrics, provenance, applicability and controlled graph recursion.

  • Local-first
  • Nea Polis meta-kernel
View case study
Knowledge SystemsExperimental

ROS.NIS / News Intelligence

Time-critical ROS domain module for events, discourse, market expectations, representation signals and separate integrity and observability analysis.

  • Local-first
  • ROS domain module
View case study

Applied in

AssessmentAssessment

Climate Assessment Europe 2026

European climate assessment covering regional climate and water impacts, societal costs and the Overshoot → Peak → Decline relationship.

vSite V6.3 · Research v3.2mixed
Open research
AssessmentWorking paper

Ling_Evo rev1.6

A thought-laboratory corpus of 14 German-language essays originating in 2024 was transformed using ROS into hypotheses, evidence runs, negative findings and explicit validation statuses.

v1.6mixed
Open research
Technical referenceWorking paper

Trieste–Opicina

A technical reference on interactions among topography, vehicle engineering, operations, urban development and the lifecycle of unusual transport infrastructure.

v2.1mixed
Open research