Catalogue and map
The questions, their sources and their limits; a globe and a detailed map of the measurement places.
DOCUMENTATION
GeoOracle studies questions about climate, water and agriculture that a public source can settle: a rule written in advance, a value read on a published source at a set date, and a file that explains the outcome. These pages describe the demonstrator: what it does and how to read it.
Demonstrator in development; nothing is in production. The records are proposals in play points, not open markets. The catalogue is a research file, not primary evidence.
Three places to start. Nothing to install.
Pick a question and read its status, its source, its threshold and its limits.
See where each measurement is made, on the globe or on the detailed map.
The observatory shows the rule, the provenance and the limits of each question.
A complete demonstrator, not yet a service.
On 16 November 2026, GeoOracle is a demonstrator. Nothing is in production.
The questions, their sources and their limits; a globe and a detailed map of the measurement places.
One file per question: rule, provenance, export and local drafts.
Each component has its tests; they validate neither a scientific result nor profitability.
The data exists. The decision still needs a basis.
Gathering sources, finding the right version and understanding exceptions take time. GeoOracle aims to provide a shared file that anyone can examine and challenge.
Support recurring decisions and reduce verification work.
Track commitments with understandable criteria and accessible documents.
Reuse rules and traceable outcomes in their own applications.
Needs to validate with their users
From question to evidence file.
One approach to know what was read, how and with which evidence.
Define the scope, period and expected outcome.
Agree on the threshold, sources and quality criteria.
Archive the data used and make the calculation reproducible.
Provide a file that can be challenged. If evidence is missing, leave it unresolved.
A place, a period, an expected outcome.
An official publication, readable without an account.
Variable, unit, threshold and reference, written before the observation.
On the set date, the published file is archived with its fingerprint.
Yes, no, or unresolved when the evidence is missing.
The file stays open to examination.
Every question has a place.
A station, a reservoir, an ocean region: the measurement is located, and that place is shown as it is defined.
A point marks a published station; an extent is an indicative window. A country outline helps to read the map; it bounds no measurement.
Plans, aerial photographs and satellite mosaics come from open sources, credited with their licence.
A location plan shows where the measurement is made, not its value. The date of the background is not the date of the question.
A market is only worth its resolution.
A prediction market gathers opinions on a question whose outcome will be known. That outcome still has to be measurable, dated and open to examination by everyone.
The question, the rule, the source and the archived evidence: enough to decide without an opaque arbiter.
No prices, no orders, no wallet. Trials use play points.
No market is open. Any trial stays a simulation in play points.
Four fields to learn from.
Pilots studied through historical archives. Their purpose: test the rules and their limits.
Check a threshold at a station, taking data quality into account.
Calculate a rainfall index over a defined period, without inferring crop loss.
Compare an official national statistic with an explicit threshold.
Check for specific documents, without treating their presence as proof of impact.
Prove the value. Then open the infrastructure.
Reproduce results, handle exceptions and know when to withhold a conclusion.
Measure time saved and errors avoided against existing methods.
Identify verification useful and frequent enough for a first customer to fund it.
Our direction: a shared format across applications and multiple verification operators. Prediction markets are a possible application of this infrastructure.
API, archiving and resolution monitoring. A business hypothesis to validate through the first use cases.
Better not to conclude than to conclude wrongly.
Six principles guide every question, every rule and every screen.
Without a public source that publishes the value, the question is classed as not measurable, with the reason.
Threshold, unit and freeze date are written before the observation. A historical replay is stated as such.
Missing, doubtful or late evidence settles nothing.
Images, plans and outlines locate; only the archive of the source resolves.
Each record says what was read, what was not, and what may change.
Play points only, with no affiliation to the organisations cited. Value is proven first on use cases.
Explore the questionsOpen the observatoryLocate the questions on the map