RESEARCH · PROGRAM AND SOURCES

Research questions
and evidence.

GenOS studies how agents can preserve state, work together, and learn from results under supervision. Each hypothesis needs a matching protocol and measured outcome before it can be treated as supported.

Browse all documentation ↗Read ADR decisions ↗

SCOPE

A summary of the material currently documented in GenOS, not an exhaustive review of external scientific literature. Prototypes, hypotheses, and measured results retain distinct statuses.

RESEARCH AREAS

Six groups of
open questions.

Each source document details its scope, technical choices, and limits. Experimental results are collected separately on the benchmarks page.

01 / RESEARCH · RUNNING

Runtime, state, and continuity

How can agent work be resumed and inspected?

GenOS explores versioned state, snapshots, counterfactual branches, and recovery. Decisions retain their provenance so results can be compared, challenged, or resumed.

EXTERNAL SCIENCE: Convergent replicated state and versioned histories — Conflict-free replicated data types (Shapiro et al., 2011)
GENOS HYPOTHESIS: Concurrent member writes can converge on shared versioned state under explicit merge operations and invariant checks.
GENOS EVIDENCE: Follow the chain →

Agent runtime Counterfactual workspaces Orchestration and branches

02 / RESEARCH · PROTOCOL

Orchestration and topologies

What organization fits a group of collaborating agents?

The work defines eight topologies, their contracts, capabilities, and limits: specialist teams, populations, communities, shared state, and execution across comparison worlds. Their presence in the runtime does not imply equal maturity.

EXTERNAL SCIENCE: Distributed scheduling and organizational coordination
GENOS HYPOTHESIS: Each topology answers a falsifiable question — e.g. three heterogeneous trajectories beat one single trajectory net of cost — testable per topology.
GENOS EVIDENCE: Follow the chain →

Topology and capability contract Topology index Agent communication

03 / RESEARCH · INCONCLUSIVE

Memory, learning, and plasticity

How can experience influence later decisions?

This research area covers episodic and semantic memory, retrieval, lessons, plasticity, and consolidation. Effects observed in controlled scenarios remain distinct from quality measured on real tasks.

EXTERNAL SCIENCE: Human episodic and semantic memory systems
GENOS HYPOTHESIS: Retrieval that combines similarity, recency, credibility, and success signals may improve long-conversation factual recall over a same-model raw-context baseline. The published LoCoMo result reports absolute F1, without that controlled comparison.
GENOS EVIDENCE: Follow the chain →

Memory and learning Neurobiology and plasticity Autobiographical memory

04 / RESEARCH · PROTOCOL

Epistemology and evidence

What justifies a conclusion or promotion?

GenOS treats outputs as results to examine. The work covers uncertainty, abstention, provenance, independent evidence, and metric limitations, separating execution success from the validity of a claim.

EXTERNAL SCIENCE: Probabilistic forecast scoring — Brier score (Brier, 1950)
GENOS HYPOTHESIS: Deliberation with explicit ballots and Brier-scored calibration improves error detection instead of merely producing consensus.
GENOS EVIDENCE: Follow the chain →

Epistemology and evidence Product completeness and status Quality and evidence

05 / RESEARCH · INCONCLUSIVE

Evolution and verified development

How can an agent evolve without confusing change with progress?

GVX and evolutionary research examine transformations, mutations, lineages, experimental nurseries, and promotion conditions. They emphasize external measurements, replication, and safeguards against circular self-evaluation.

EXTERNAL SCIENCE: Metapopulation ecology: local extinction and recolonization — Levins metapopulation model (Levins, 1969)
GENOS HYPOTHESIS: Semi-independent groups under extinction pressure maintain capabilities through migration and recolonization.
GENOS EVIDENCE: Follow the chain →

Verified development (GVX)GVX experimental nursery Evolutionary research on selfhood

06 / RESEARCH · HYPOTHESIS

Biomimetic systems and cognition

Which biological ideas can become testable computational mechanisms?

The program examines memory, homeostasis, communication, perception, immunity, and collective coordination as engineering models. Analogies organize mechanisms; they do not prove biological properties or consciousness.

EXTERNAL SCIENCE: Physiology, immunology, and collective behavior as design analogies
GENOS HYPOTHESIS: Biological organization patterns translate into bounded runtime operations with explicit limits — inspiration first, validation per mechanism.
GENOS EVIDENCE: Follow the chain →

Computational biology Epistemic immune system Vital agent organs

PRIMARY REFERENCES · MECHANISMS

Sources for the
mechanisms.

These primary sources describe selected scientific mechanisms used as design analogies. They do not test or validate GenOS code.

Episodic vs. semantic memory · 1972

Episodic and semantic memory

Endel Tulving · Organization of Memory, pp. 381–403

Research questionsand evidence.

Six groups ofopen questions.

Runtime, state, and continuity

Orchestration and topologies

Memory, learning, and plasticity

Epistemology and evidence

Evolution and verified development

Biomimetic systems and cognition

Sources for themechanisms.

Episodic and semantic memory

The Adaptive Nature of Human Categorization

Neural networks and physical systems with emergent collective computational abilities

Synaptic modifications in cultured hippocampal neurons: dependence on spike timing, synaptic strength, and postsynaptic cell type

Synaptic modifications in cultured hippocampal neurons: dependence on spike timing, synaptic strength, and postsynaptic cell type

Verification of forecasts expressed in terms of probability

A logical calculus of the ideas immanent in nervous activity

Optimal foraging, the marginal value theorem

The Use and Abuse of Vegetational Concepts and Terms

Concluding Remarks

Some Demographic and Genetic Consequences of Environmental Heterogeneity for Biological Control

La reconstruction du nid et les coordinations interindividuelles chez Bellicositermes natalensis et Cubitermes sp. La théorie de la stigmergie : essai d'interprétation du comportement des termites constructeurs

Flocks, herds, and schools: A distributed behavioral model

The logic of animal conflict

Intelligence without representation

Epigenetic Control Systems

Molecular Structure of Nucleic Acids: A Structure for Deoxyribose Nucleic Acid

Time, Clocks, and the Ordering of Events in a Distributed System

The Byzantine Generals Problem

An axiomatic basis for computer programming

The protection of information in computer systems

Design of fault-tolerant computers

A mathematical theory of communication

Distributed snapshots: determining global states of distributed systems

The chemical basis of morphogenesis

The wisdom of the body

The problem of anabiosis or latent life: history and current concept

The origin and behavior of mutable loci in maize

The clonal evolution of tumor cell populations

The source code control system

The transaction concept: virtues and limitations

Architecture of SQLite

The design of experiments

Is justified true belief knowledge?

Acquisition of cognitive skill

Towards a cognitive neuroscience of consciousness: basic evidence and a workspace framework

A modification of Jerne's theory of antibody production using the concept of clonal selection

What we measurestays contextual.

Research questions
and evidence.

Six groups of
open questions.

Sources for the
mechanisms.

What we measure
stays contextual.