SystemsEvidenceAccess

smJVUIAFv txaAkJFeraK

Three stages, same standard at every one. Scope determines depth, not company size.

First Light

Dingir world model · evidence-first · cross-domain intelligence
You bring
  • What you need looked at, an app, a claim, a system, a competitor, a vendor promise, a pitch deck
  • What's riding on it, a signature, an investment, a decision, a story you're about to act on or publish
  • Anything already public about it, we don't start from zero
We
  • Run a non-classical review at the root level, through Dingir, our own world model, with our core human team and agent swarm
  • Contextualise everything found and bring it into one coherent shape, never hand it back as only a raw finding list
  • Trace every claim back to real evidence, not a summary
You receive
  • A clear verdict or findings report, not a plausible-sounding guess
  • The evidence trail behind it, the same sourcing we'd stand behind ourselves
  • Something you can sign off on, invest on, publish on, or act on
Delivered within 14 calendar days.

Didn't find the right fit? Talk to us - get a custom quote →

Engagement Journey

what happens after you start

Five stages, from a single observation to a continuously improving intelligence model. The depth changes with the question; the discipline stays the same. Every stage stays traceable end to end, whitebox by design, so any action can be backtraced to the observation that produced it.

1
Ingest

We start with what you trust us to handle: a signal, a claim, a system, or a question - from your own environment, public sources, or a domain we already monitor.

Nothing gets touched until we've read the actual thing, not a README, not a summary someone else wrote about it. The starting point can be small. What we check it against is not.

2
Normalize

We resolve entities, map relationships, attach timestamps and provenance, and preserve what remains genuinely uncertain instead of guessing. Different vocabularies, different systems, different languages become comparable without being flattened into sameness.

Nothing is accepted without a source. Nothing unknown gets silently filled in.

3
Trace

This is where three purpose-built engines do the actual work: TemporalEngine compares states over time, NetworkEngine follows dependencies to find bridges and bottlenecks, PatternEngine tests whether what we're seeing matches something we've already found elsewhere.

The question was never only what changed - it's what else it touches.

4
Emit

A meaningful change becomes an intelligence event: evidence, confidence, contradiction state, and every domain it affects, written into EvidenceGraph as one traceable object, not buried in a paragraph you have to take our word for.

Not a verdict yet. Something you can still investigate and challenge.

5
Learn

This is where it reaches you: the finding, the evidence behind it, and what it changes for you - one deliverable you can actually act on.

It doesn't end there. The same evidence stays in EvidenceGraph, so the next question about the same subject starts from what we already know, not from zero - every engagement makes the next one faster.

© 2026 RFI-IRFOS  ·  Graz, Austria  ·  GISA 39261441  ·  UID ATU83405245

Human rights are not subject to negotiation.

Foundation

Intelligence

World Model

Applications

Repositories

Crates (crates.io)

all packages →

this is a useless cookie banner. it's just here to look like one * we don't use cookies, so there's nothing to consent to. don't let anyone tell you otherwise.two buttons, one closes this and throws some confetti. the other literally does nothing.