Research graphs
A research graph is a navigable map of a topic that shows how the people, organisations, and events connect, grades every claim as confirmed, alleged, or rumoured, and points you at what to research next. I build them as an Obsidian knowledge graph you can open, search, and hand to a colleague. Every connection is an explicit claim with a source and a grade, the graph flags the central things it has not investigated yet, and it stops when it runs out of real findings instead of padding the answer. The result is a structured artefact you can audit, not a chat transcript you have to trust on faith.
Who it is for
- Business owners who need a subject researched properly before a decision rides on it, and want to see which parts are verified and which are still guesswork.
- Agencies whose client asked for AI research and who need the depth done right, white-labelled or behind the scenes.
- Investors and analysts who need the people, money, and events on a topic mapped and graded, with source conflicts written up rather than smoothed over.
What you get
An evidence-graded, audited, navigable knowledge graph that keeps growing. Every claim carries its grade. The graph surfaces its own blind spots and ranks them by how much they matter. A final audit re-reads every note against its sources, and a health scorecard has to read clean before the graph is called done. Run another round later and it extends from where it stopped.
Give me a topic, three to five seed entities, and your sources. I run the machine.
Read the full method
This is the short version. The full method shows exactly how it works: the evidence grades, the blind-spot detection, the convergence rule that decides when to stop, and the audit that closes a graph out. The same machine is already running in two live products, SignalTrace and JobAbroad.
To start, email me at info@devai.co.za and tell me what you are trying to understand.
