Built to make investment judgment inspectable.
Investment theses need memory.
Crene gives live institutional views a durable structure: assumptions, disagreement, changes, and rethink triggers in one record.
Investment teams have models, notes, data feeds, expert calls, and debate. What they often lack is a living record of the assumptions underneath the view.
Crene frames each thesis with a horizon, decomposes the conditions that would decide it, tracks model disagreement, and preserves what changed.
The aim is not to produce a trade trigger. It is to make thesis review inspectable before a PM, risk, CIO, or investment committee discussion.
A thesis should have a memory.
The system should remember what the team believed, why it believed it, and what changed.
Disagreement should be visible.
Model disagreement, assumption pressure, and missing evidence should be surfaced before review, not buried after the fact.
Calibration should be disclosed.
Where outcomes resolve cleanly, Crene scores the record and shows the misses.
Maps are not predictions.
Structural maps show what a thesis depends on. Scored forecasts stay separate unless the outcome is cleanly resolvable.
Stephen spent nearly a decade in finance, including roles at Goldman Sachs and Crédit Agricole, before teaching himself full-stack software engineering.
Crene grew out of a problem he saw in institutional decision making: teams could track prices, models, notes, and meetings, but the assumptions underneath a thesis were rarely preserved as a system. When the world changed, it was hard to see which part of the view had actually changed with it.
Stephen builds across the full Crene stack, from data collection and model scoring to backend infrastructure and frontend product surfaces.
I built Crene because investment teams have models, notes, decks, data feeds, and expert calls, but no living system for the assumptions underneath a thesis. Crene is the product I wanted before a thesis review: a way to see what the view depends on, where confidence is moving, where models disagree, and when the team should rethink.
Crene is useful only if the record can be inspected. We publish what can be scored, separate maps from forecasts, and avoid claims the data cannot support.
Calibration disclosed.
Where outcomes resolve cleanly, Crene scores the record in tiers and includes the misses.
Maps stay separate.
Structural maps show what a thesis depends on. Scored forecasts are kept separate unless resolution is clean.
No false precision.
Crene does not claim to price tail events or the genuinely unmodelable.