How it works
A persistent scientific campaign — not a chat answer
Camile Scientific Intelligence runs investigations, not replies. Each campaign is a durable object: it accumulates research state, links every claim to evidence, and advances only through explicit scientific gates.
The campaign loop
The campaign is the fundamental unit of the platform. From a research question to a traceable result, every stage is explicit, versioned and auditable.
1 · Research question
Every campaign starts from a question stated before the work begins — with success criteria defined up front.
2 · Research state
The campaign maintains a living state: what is known, unknown, assumed and contested — versioned over time.
3 · Evidence
Findings enter as evidence with provenance. A claim without a traceable source does not move the state.
4 · Hypotheses
Competing hypotheses are tracked explicitly. The system reasons with rivals — not a single favorite.
5 · Predictions
Each hypothesis yields falsifiable predictions, registered before the data arrives.
6 · Experiments
Computational experiments run in a deterministic sandbox. Failures and null results are first-class outcomes.
7 · Verification
Predictions resolve against outcomes — verified, refuted or inconclusive. Every resolution feeds calibration.
8 · Results
Converged results carry full lineage: from question to artifact, reproducible end to end.
9 · Publication
Outputs are packaged with provenance and released only under explicit human authorization.
What makes it different
- Persistent campaigns: investigations that run for days or weeks — not conversations that end in minutes.
- Evidence before assertion: every statement about the world must carry a pointer to how it is known.
- Pre-registered falsifiability: predictions are locked before data collection, so results cannot be quietly reinterpreted.
- Calibrated uncertainty: the platform reports what it does not know, with confidence that is measured — not assumed.
- Reproducibility as a product feature: any result can be replayed and independently re-run.
- Human oversight at every gate: autonomy inside the loop, authorization at the boundaries.
See a campaign end to end
The research workspace shows the full loop — state, evidence, predictions, experiments and verification — on every campaign.