The part nobody else has: the ledger.
Thesis scores every locked Thesis against what actually happened — and feeds the result back into the context snapshot. Judgment becomes a measurable, improving, compounding asset.
| Thesis | Locked | Conf | Outcome | Brier | Status |
|---|---|---|---|---|---|
| SSO for mid-market | Feb 2026 | 80% | Hit (expansion +19%) | 0.04 | HIT |
| Mobile companion app | Jan 2026 | 70% | Missed (DAU 6% vs 25% target) | 0.49 | MISS |
| Usage-based pricing tier | Mar 2026 | 60% | Hit | 0.16 | HIT |
| Zapier deep integration | Mar 2026 | 75% | Partial (adoption hit, retention flat) | 0.31 | PARTIAL |
| Onboarding checklist v2 | Apr 2026 | 85% | Hit | 0.02 | HIT |
| AI summary emails | Apr 2026 | 65% | Missed (opt-out 34%) | 0.42 | MISS |
| Template marketplace | May 2026 | 55% | Killed at exit condition, week 3 | 0.2 | KILLED |
| Scalable roles & audit trail (TH-014) | Jul 2026 | 72% | Pending | — | OPEN |
Three assets no competitor can copy.
This signature is why Act 2 flagged "AI compose" as drift — and why Act 3 registered 72%, not 85%.
Detection is reweighted by this record — Finding 1's "high confidence" was computed from Meridian's own history, not a generic rulebook. Convergence isn't a heuristic here; it's an empirically weighted model per org.
Priors from the Thesis network — workflow-automation category.
Before your team locks a bet, Thesis already knows how bets like it have gone across the network.
The context itself learns.
Decisions update the snapshot; the snapshot shapes the next decisions. Over quarters, the loop compounds.
Thesis can score decisions you made before Thesis existed — point it at 12 months of Linear and analytics history and get your ledger on day one.
Every quarter, Meridian's ledger gets harder to live without — and impossible for a competitor to copy.