Anyone can tell you what happened. Almost nobody can tell you what you believed beforehand, and why, because that only exists if something was writing it down before the answer was known.
Public Market Memory is the record: what companies said, what they promised, what they reported, what changed. It is all public, so it can be reconstructed, and it is, which is why LyraMind is useful on day one rather than after a year of accumulation. Onboard an issuer and see its commitment history immediately, including what quietly stopped being said.
Private Institutional Memory is yours: what your firm believed, who changed their mind, with what conviction, on what evidence. This genuinely cannot be reconstructed. It begins the day LyraMind arrives.
The pre-outcome state is the whole asset. If it could be edited after the fact, it would record what the firm wishes it had believed. Nothing is overwritten: an assumption moving from intact to weakening appends a transition, and the prior state stays readable forever.
A decision made in March cannot be replayed against April's evidence. If it could, every calibration statistic built on the archive would measure hindsight rather than judgement, and the lead-time figure in particular would silently become a measure of how long you waited rather than how early the warning appeared.
Every one of those carries its sample size, because a calibration claim without an n beside it is a claim nobody can weigh.