Search and RAG answer “what content should I retrieve?” Idea Network continues into “how did this work evolve, where are the evidence boundaries, and where should the investigation go next?”
Different systems do different jobs. Search provides an entry point, RAG brings relevant content into a generation, and knowledge graphs represent entity relations. Idea Network connects these capabilities into an evidence-bearing, stateful research knowledge environment.
Four capabilities, not four synonyms
| Capability | Core question | Typical output | Cross-session state |
|---|---|---|---|
| Literature search | Which papers may be relevant? | Ranked papers and abstracts | Usually limited |
| RAG | Which passages should support this answer? | Retrieved chunks and generated response | Depends on external design |
| Knowledge graph | How are entities related? | Nodes, edges and attributes | Stores factual relations |
| Idea Network | How did the research evolve, what are the boundaries, and where next? | Relation paths, source evidence, unseen/blind spots and research sessions | Preserves exploration and redirection |
What Idea Network adds
From similarity to why something is related
Similarity produces candidates, not proof of a research relation. Idea Network combines citations, authors, concepts, ontologies, methods and time so an agent can explain why a paper belongs on a path.
From retrieval results to evidence paths
An agent gets compact information first, then drills into abstracts and source passages as needed. Provenance remains attached at each layer, reducing irrelevant context.
From one answer to a research session
Research continues after a generation. Idea Network preserves what has been viewed, excluded, confirmed and redirected outside the model, allowing the next agent or team member to resume.
From “not found” to honest boundaries
The system distinguishes graph-supported, not found in the graph and graph blind spot. When the current scope yields no evidence, the system reports only that it was not found; when coverage or retrieval paths are insufficient, the blind spot is explicit. Provenance and uncertainty remain in the workflow instead of disappearing behind fluent language.
Where Idea Interflowing sits
Idea Network is the knowledge environment. Idea Interflowing is the pre-experiment research-decision SaaS built on it. Idea Creator follows relations and evidence to surface structural gaps and form proposals. Idea Supervisor reviews papers or completed work against target-venue criteria and links each judgment back to evidence.