Merlin — the OIS visible brain.
Merlin is the continuous intelligence worker inside OIS. It observes real sources, runs the reasoning engines, and turns what it finds into evidence-backed Intelligence Objects with recommendations for a person to decide. It works around the clock — and it never decides alone.
What Merlin is
Merlin is where OIS stops being an idea and starts doing the work. It is a standing worker made of many small, single-purpose agents: they read sources, corroborate claims, run the OIS engines, and assemble each finding into an Intelligence Object — the object with its sources, claims, entities, relationships and evidence in one place. Merlin recommends; a person decides.
How Merlin works
Every step is traceable to its evidence. Merlin proposes; people decide — nothing is published or acted on without a human gate.
What Merlin does
Reads and corroborates
Follows sources, extracts claims, and checks them against independent origins before trusting them.
Runs the engines
Routes each task to the OIS engine built for it — relationships, memory, causes, foresight, legitimacy — instead of one undifferentiated answer.
Builds Intelligence Objects
Assembles each finding into a structured record: sources, claims, entities, relationships and evidence together.
Proves its work
Every workflow ends in an evidence-backed object; where it cannot corroborate, it says so.
Surfaces recommendations
Proposes the next step and leaves the decision to a person.
Watches for gaps
Flags what is stale, missing or unverified — instead of hiding it behind confident output.
The engines Merlin runs
Merlin operates the OIS reasoning engines as living code, not a diagram. Seven that were once design-only now run continuously: Nexus (entities and relationships), Memoria (structured memory and precedent), Praedicta (possible developments — not deterministic prediction), Sentiment (tone and stance from evidence), Discovery (finding and comparing sources), Drivers (the causes behind an event) and Insight (turning it into a usable brief). Each persists its own real output that can be inspected — not a claim on a slide.
A worker, not an oracle
Merlin can be wrong — which is exactly why every output is evidence-backed and gated by a person. It does not accept its own generated text as fact: a signal must trace to evidence the trust layer can point to. Where something cannot be corroborated or measured, Merlin marks it as not verified, never dressed up as insight.
Proven in a real newsroom
Merlin runs live inside Belgium Impulse, a real editorial environment with real consequences. There it works as a standing team of specialised agents — observing sources, running the engines and drafting under editorial gates — producing genuine editorial intelligence continuously, under human authority. This is a demonstration that the ideas work in practice; it is evidence, not a universal guarantee for every organisation.
What Merlin does today (measured)
- A standing team of specialised agents, each with one job
- Runs the seven OIS engines continuously, on a schedule
- Maintains a live entity graph of the relationships it finds
- Corroborates claims against independent sources before trusting them
- Drafts editorial work for Belgium Impulse under human gates
- Ends every workflow in an evidence-backed Intelligence Object
- Marks what is stale, missing or unverified as not measured
See your organisation clearly.
Start with a read-only OISA audit — what is known, what is missing and what to decide.