The InsightNodes Architecture
The intelligence system behind the connections.
Structural intelligence, investigated by Agentic AI.
InsightNodes combines a persistent relationship graph, sourced evidence, deterministic analysis and Agentic AI to investigate how companies, technologies, capital, infrastructure and markets connect — and what those relationships actually mean.
What's Underneath It
Five layers. One structural intelligence system.
The Graph
Persistent structural relationships between companies, technologies, capital, people, infrastructure and policy.
Evidence
Sourced relationships, dated milestones and full provenance — traced back to a filing, announcement or clearly labeled source.
Analysis
Multi-hop traversal, derived structural measures, temporal change and deterministic scenario propagation.
Agentic AI
Investigates a research objective across the graph, the evidence and the relationship history — not a single-pass summary.
Challenge & Verify
Tests findings for gaps, contradictions, unsupported claims and missing context before anything is presented as an answer.
How The Graph Works
See the structure.
Below is the actual InsightNodes relationship graph, centered on NVIDIA — live data, not a mockup. This is the structural layer Agentic AI investigates, not simply an interactive visualization: every node and edge here is exactly what the agents traverse, cite and challenge when they answer a research objective.
Connection type
Click any node to open its entity page. Scroll to zoom, drag to pan, or use the fullscreen toggle to open the full map.
This is a first-degree view. The full entity page lets you go 2, 3 or 5 hops deep.
NVIDIA depends on TSMC for advanced semiconductor manufacturing.
which depends on↓ASMLTSMC depends on ASML for the lithography equipment that manufacturing requires.
A real, sourced two-hop chain — not an illustration.
Trace the network →How Agentic AI Uses It
See the intelligence behind the structure.
A real, captured InsightNodes research investigation — not a mock. This shows how Agentic AI moves through the five layers above on an actual research objective.
Research objective
“Map the direct and indirect relationships connecting NVIDIA with OpenAI.”
Direct answer
NVIDIA and OpenAI are directly linked through both a supply and an investment relationship: NVIDIA is OpenAI's major GPU supplier, committed to a 10-gigawatt systems deployment with up to $100 billion in related investment, and was one of three co-leads (alongside Amazon and SoftBank) in OpenAI's ~$110 billion 2026 funding round at an $840–852 billion valuation. Indirectly, they are connected through Microsoft (which invests in OpenAI and depends on NVIDIA GPUs for Azure) and through TSMC, a shared upstream manufacturing dependency for NVIDIA and competitor-turned-OpenAI-supplier AMD.
Key structural findings
- OpenAI depends on NVIDIA for GPU supply (model training/inference); NVIDIA also invested $30 billion as one of three co-leads in OpenAI's 2026 funding round alongside Amazon ($50B) and SoftBank ($30B), at an $840–852 billion valuation.
- NVIDIA and OpenAI announced a September 2025 letter of intent for at least 10 gigawatts of NVIDIA systems deployment, with NVIDIA intending to invest up to $100 billion as systems deploy; a separate August 2026 arrangement has NVIDIA backing up to $105 billion in financing for an OpenAI Ohio data center.
- Microsoft has invested tens of billions in OpenAI and is its largest cloud provider, while Microsoft separately depends on NVIDIA GPUs to build Azure AI infrastructure — creating an indirect NVIDIA-OpenAI link through a shared major partner.
Captured from an actual InsightNodes research run.
From AI assistant to intelligence system
Five specialist AI agents. One connected answer.
InsightNodes is building toward five specialist AI agents, coordinated around a common structural intelligence layer — each focused on a critical stage of the research process: exploring the knowledge graph, gathering and verifying evidence, challenging findings, and synthesizing a clear, structured answer.
1
Understand
Research Agent
Interprets your question, identifies key entities and defines the research plan.
2
Investigate
Graph Agent
Explores the knowledge graph to uncover relevant relationships and context.
3
Verify
Evidence Agent
Retrieves and evaluates evidence from trusted sources.
4
Challenge
Analysis Agent
Tests for gaps, contradictions and alternative explanations.
5
Synthesize
Synthesis Agent
Combines findings into a clear, structured answer with actionable insights.
Connected
Intelligence.
The agents are specialized. The graph is shared.
Each agent focuses on one stage of investigation, but all of them operate over the same persistent structural graph and evidence base — not five disconnected tools.
Today: a single research orchestrator performs these stages — graph investigation, evidence retrieval, challenge, synthesis — within one bounded research loop. Roadmap: separating those stages into the five specialized agents above.
What The System Computes
From information to intelligence
What InsightNodes computes that nobody else does
Every row below is live on NVIDIA's actual page right now — click any one to see it for real.
| Understand | |
| Relationship Map | An interactive, clickable map of how an entity connects — ownership, dependencies, regulation, technology and more, several hops out. |
| Confirmed Dependency Edges | Connections in the Relationship Map get a checkmark when an earnings-call mention or patent citation independently corroborates that edge. |
| Relationship History | How a connection between two entities evolved over time, not just its current state. |
| Strategic Importance & Confidence | Every entity is rated Critical to Emerging, with a transparent confidence score on how well-sourced the claim is. |
| Monitor | |
| Activity (Relationship Velocity) | Whether an entity's recorded connections are accelerating, cooling, or steady over the last 90 days. |
| Momentum | Combines hiring, patent filing, and insider filing trends into one directional read. |
| Emerging Signals | Forward-looking shifts in the landscape, tied to the entities they affect. |
| The Story So Far | An auto-generated timeline built purely from dated milestones and sourced updates — not AI-written, just sorted. |
| Analyze | |
| Chokepoint Score | Computed fresh from the live graph — how many other entities structurally depend on this one. |
| Correlated Activity | Flags when an entity and something it depends on (or that depends on it) both show a burst of recorded change in the same window. |
| Insider Timing Correlation | Cross-references an insider's SEC Form 4 buy/sell timing against dated milestones and relationship changes at the same company. |
| Strategic Intelligence Report | Every entity profile can be published as a clean, shareable PDF-style report. |
| Verify | |
| Milestones | A sourced, dated history of the entity's key moments — founding to present. |
| Market Intelligence | Credibly-reported but unconfirmed claims, kept structurally separate from verified evidence. |
| Earnings-Call Mentions | Literal, dated quotes pulled straight from a company's own SEC 8-K filings when they name another tracked entity. |
| Source Intelligence | A public look at how well-sourced the graph is, tier by tier. |
| Decide | |
| Insider Filing Activity | Tracks Form 4 filing frequency straight from SEC EDGAR. |
| Federal Contract Disclosures | Real federal award data from USAspending.gov — agency, dollar amount, and award ID, updated monthly. |
| Compare Workspace | Put two entities side by side — which holds the stronger strategic position, and why, built from real graph metadata. |
| Scenario Simulator | Model a shock to one entity and trace, deterministically, how it propagates through the graph. |
| Portfolio Intelligence | See a group of holdings as one graph — shared dependencies, competitive overlap, and second-order exposure a single entity view would miss. |
What A User Can Do
Five ways to use the graph.
Start with a question. Choose the mode that matches the decision, then inspect the relationships, timing and evidence behind the answer.
01
Discover
What connects?
Reveal direct and indirect relationships around a company, technology or market. Use relationship and temporal intelligence to see what formed the connection and how it changed.
02
Trace
What depends on what?
Follow suppliers, capital, infrastructure and second-order exposure. Dependency intelligence shows where a chain is concentrated or fragile.
03
Compare
How do structures differ?
Compare companies, ecosystems or holdings by shared exposure, concentration and structural position. Portfolio intelligence makes overlap visible.
04
Simulate
What happens if this changes?
Run a scenario through the graph to test how a disruption, constraint or strategic move could propagate across connected entities.
05
Investigate
Where is the hidden bottleneck?
Start from new evidence or a suspected constraint, then trace where the effect may travel. Bottleneck intelligence surfaces the nodes that can reshape the system.
Evidence / Trust
AI that can show its work.
Every answer traces back: source facts, graph facts, derived measures and interpretation are kept separate — never blended into an unlabeled claim.
The core chain
Evidence → Structure → Implication → Decision
Scale of the Graph
The graph, at scale.
8,216+
Entities
33,363+
Relationships
36,986+
Evidence Citations
3,457+
Dated Milestones
116+
Countries
56+
Verticals
The graph compounds.
Every validated relationship creates new paths to investigate — the value isn't the record count, it's the structure connecting them.
Don't take our word for it. Explore the structure.
The fastest way to understand InsightNodes is to use it — see the relationships, trace the dependencies, run a scenario, or let Agentic AI investigate a question for you.