Provenance belongs in the answer.
Finding information is easy. Knowing whether to trust it is harder. We are building discovery that puts sources, confidence, and agreement in front of the reader instead of burying them in machinery.
Working implementation with internal verification.
The problem
Finding information has never been easier, and knowing whether to believe it has rarely been harder. Search returns ranked lists without saying why. Generative answers return fluent text that reads identically whether it is well supported or barely supported at all.
So the hard judgement has moved to the reader. How good the source is, whether anyone else agrees, how much confidence an answer deserves. All of it lands at the moment the reader has the least context to decide.
Trust is treated as invisible infrastructure. We think it belongs in the interface, where a person can actually use it.
What we are doing
Answers that carry their sources
Responses are grounded in retrieved material and cite it, so a claim can be followed back to where it came from.
Confidence stated, not implied
Where support is thin or sources disagree, the interface says so, instead of presenting every answer with identical assurance.
Consensus made visible
Whether sources agree is surfaced as part of the result, because agreement and disagreement are information in their own right.
Built for decisions, not clicks
Comparison and evaluation are treated as first-class intents, because people are usually trying to decide something.
Technology in play
Proof of work
What we can evidence today, with the limits of that evidence stated alongside it.
A working search stack has been built, with a live primary retrieval path.
Verified internally. Index design, source pipeline, and providers are not published.
Citation-backed answers, confidence states, consensus labelling, and comparison intent have been implemented and internally verified.
The capabilities are real; the weighting and scoring behind them are proprietary.
Safety and abuse-resistance components are part of the system rather than a later addition.
Thresholds and detection behaviour are deliberately not described.
Designed responsibly
Transparency without a manipulation manual
We show people why to trust a result without publishing the mechanics that would let the system be gamed.
Attribution to the source
The system is designed to point back to publishers rather than absorb their work and present it as its own.
Safety as architecture
Abuse resistance is part of the design rather than a filter applied at the end.
Why it matters
Synthetic content is now cheap and abundant. The scarce thing is not information; it is warranted confidence in information.
A discovery system that makes provenance and confidence visible changes what people can reasonably be held responsible for knowing. Every serious institutional decision rests on that.
Explore NOVA
- Intelligent SystemsGovernance that runs at the moment of action, not after the incident.
- Financial InfrastructureIdentity, movement, risk, and the ledger designed as one system.
- Work & Human SystemsLearning, hiring, and employment held in a single continuous record.
- Health SystemsReception, records, devices, and follow-up moving as one flow.
- Public SystemsNational-scale systems that integrate without displacing institutional authority.