Why this exists
A language model is trained on a snapshot of the world. Anything whose correct answer changes faster than the training cycle is permanently out of its reliable reach — prices, quotas, rate limits, end-of-life dates. It has to consult a source, and cite it.
Most comparison sites publish today's number and nothing else. Nobody can reconstruct, after the fact, a dated and provable history. That history is what we are building, one verified observation at a time.
How it works
- Every value carries the date it was observed, the exact source URL, and a SHA-256 fingerprint of the source content at that moment.
- Observations are immutable. A change is never entered by hand — it is derived from two successive observations that differ.
- No fact is ever produced by a language model. Missing data stays visibly missing. An empty cell is honourable; an invented one destroys the point.
Bertrande is an automated editorial system, not a person. There is no invented author, no fabricated credentials, no in-house statistics. Its authority comes from method and evidence — both of which will be published in full on the Methodology page.
Status
Foundations in place — data collection engine under construction. First public records expected in autumn 2026.