Analyses.

The public archive, dated. Every analysis is a timestamped record: the probability, the reference class it anchors to, and the criteria that would prove it wrong.

2026

Ukraine drone campaign against Russian logistics — interstate war duration
full analysis · graded claims · duration-conditioned prior · adversarial peer review70% · likely · war still active at 12 months OPEN
23 Jul 2026 St Petersburg drone strikes — interstate war duration
full analysis · graded claims · the argument against · adversarial peer review55% · likely · war still active at 12 months OPEN
17 Jul 2026 Lebanon ceasefire — full interactive analysis
full analysis · directional-call case study · outcome note on page40% · narrative holds UNKNOWN(pre-registry)
16 Apr 2026 Iran ceasefire — the 21-minute call
case study · timestamped provenance · includes what we got wrong75% · de-escalation RESOLVED
7 Apr 2026 Iran ceasefire — full analysis (redacted)
full analysis · redacted for publication75% · de-escalation RESOLVED
7 Apr 2026

Badges: OPEN horizon not reached · RESOLVED adjudicated against pre-registered criteria · UNKNOWN(pre-registry) published before the question registry existed — not adjudicated.

Why the public archive is short while the system is not. VERA has produced hundreds of internal analyses since April — the daily falsifier sweep walked 361 records on its first run. Public releases are held to a stricter bar: each one ships with its falsification criteria and joins the scoreable record. Publication cadence rises with the first measured cohort in August.

site updated 18 July 2026