What Prevo is.

News platforms stop at publication. Prevo treats a policy event as something with a before and an after: the announcement, the material claims inside it, the precedents that resemble it, and the outcomes measured at three, six and twelve months. Sequence is never presented as causality, and the archive gets more useful with every event that joins it.

Every material factual claim is linked to supporting evidence that meets Prevo's source requirements, down to the page and table where the figure lives. A claim the evidence cannot support is labelled unverified rather than dressed up, because a verification level the evidence does not earn is a defect and an honest "unverified" is not.

Corrections are recorded, never hidden. A claim that turns out to be wrong is marked contradicted and kept, with what was said and why it went, so the record shows its own repairs.

One AI drafts. A second, independent AI attacks every number. Code verifies what models should never assert: date arithmetic, rate pairs, window labels. A human approves every event before it publishes. Agreement between models is never treated as truth, and nothing publishes on model consensus alone.

Prevo compresses the research, not the thinking. Every analysis exposes the case for, the case against, its key assumptions and what could make it wrong; the Prevo view is always labelled interpretation, never fact. The product gets you to your own view faster. It never hands you one to repeat.

Alongside the archive sits the Singapore layer map: seventy six government spatial datasets on one canvas, from URA, HDB, SLA, LTA, JTC, NParks and five more agencies, so the boundaries that shape a site can be read together.

Questions, corrections and partnerships all start the same way: write to us. If you believe a published figure is wrong, say so; the corrections record is part of the product.