Insights / Methods
From finding papers
to seeing a direction
The bottleneck for research agents is no longer model capability alone. It is whether a trusted, traceable and stateful knowledge environment exists outside the model. These articles explain Zhiway's thesis, product boundaries and field method.
Core Questions
One article, one question
Each piece starts with a definition, then explains mechanism, boundary and application.
What is agent-native research knowledge infrastructure?
Why larger models do not automatically eliminate knowledge freshness, evidence provenance, research state or context cost.
How does Idea Network differ from search, RAG and knowledge graphs?
Search finds content and graphs represent relations. Idea Network continues into evolution, evidence boundaries and next directions.
What is Research FDE, and why must AI4S enter the field?
How general capabilities connect to private institutional data, specialist tools and tacit workflows, then become concrete tasks with explicit acceptance and procurement criteria.
Canonical Terms
One term, one stable meaning
- Idea Network
- An agent-native research knowledge environment for relations, evidence and research state.
- Idea Interflowing
- A pre-experiment research-decision SaaS built on Idea Network.
- Research Claw
- A local research-agent environment inside institutions.
- Research FDE
- The field method for adaptation, acceptance and procurement boundaries.
Let agents see how the literature connects.
Start with a trusted knowledge environment, so agents can see which directions are worth pursuing before an experiment begins.