About
Vaethra exists because the information needed to see a disruption coming is almost all public, almost all scattered, and almost never assembled by anyone without a reason to shade it.
The problem is assembly, not access
Almost everything you would need to see a crisis forming is already published. Seismic networks publish. Port authorities publish. Statistical agencies publish trade flows. Grid operators publish load. Coded conflict datasets publish incidents. Nearly none of it is secret.
What does not exist is the assembly: one place where a strait's traffic falling can be read next to who depends on that strait, next to what those parties in turn supply, with every step carrying the source it rests on. That work is unglamorous, it is mostly plumbing, and it is the difference between a headline and an early warning.
The organisations that do assemble it sell access for tens of thousands a year, which means the people most exposed to disruption — journalists, researchers, small operators, anyone in a country the market considers too small to sell to — are looking at the same headlines as everybody else.
Four commitments, in the software
These are not values on a wall. Each one is enforced somewhere in the code, and each one has cost something to keep.
Provenance travels with the record
Every figure carries the publisher it came from and the moment it arrived. Not in a footnote — attached, everywhere it appears, so any claim can be walked back to something you can check without asking us.
The accuracy record is published, including the misses
The backtest reports how often signals actually preceded confirmed events, scored against two null baselines. Where the sample is too small to support a claim it says so and prints no rate. There is no parameter anywhere that suppresses a bad result.
Estimates are labelled as estimates
Observation and model output are separated everywhere in the interface, because the fastest way to destroy an intelligence product is to let the second quietly acquire the authority of the first.
The observation layer stays free
No account, no cookies, no tracking, no paywall on the evidence. It stays free because it is also the proof: an accuracy record behind a paywall is worth nothing to anyone deciding whether to believe it.
Ziad Soliman
Vaethra is built and run by Ziad Soliman, who designed and wrote the whole of it — the collection layer, the dependency graph, the cascade model, the scoring harness and the interface.
The shape of the product follows from a single conviction that runs through all of it: a system that cannot be checked should not be believed, including this one. That is why the evidence tiers exist, why the backtest publishes its failures, why model output is fenced off from observation, and why the free layer is the evidence rather than a sample of it.
It is also why the work is unusually public. The methodology, the glossary, the coverage gaps and the limits are all documented here rather than described in a sales call, on the view that a claim nobody can inspect is a claim nobody should act on.
Enquiries — press, partnership, data licensing, or anything about the method — contact@vaethra.com.
The order of construction
Four layers, built deliberately in sequence rather than in parallel. Observation first, because everything above it inherits its errors. Consequence second, because knowing what rests on what is the part that turns an event into a forecast worth making. Prediction third, once there is a harness that can tell whether a forecast was any good. And a fourth layer studying how populations interpret all of it, because belief moves people as reliably as scarcity does.
| Layer | Question | State |
|---|---|---|
| Terminal | What is happening? | Live |
| Atropos | What does it break? | Live |
| Oracle | What happens next? | In development |
| Eschaton | What do people believe it means? | Early |