Capability Map
Capabilities
| Capability | Maturity (0-5) | Evidence | Classification |
|---|---|---|---|
Algorithm framework (libs/agency/src/lib/algorithms/) | 4 | sales-forecasting, compound-rate-tracking proven | Core |
AlgorithmMetadata export pattern | 4 | Standard across all algorithms | Core |
algorithm-constants.ts | 4 | Established pattern, extensible | Core |
| CRM data access (Drizzle repos) | 4 | 5 deals, 10 activities, 28 contacts in prod | Supporting |
| Convex agent messages (HTTP client) | 3 | Read channel working, query patterns proven | Supporting |
| Commissioning state (markdown) | 2 | Manual L-level tracking in index.md | Supporting |
| Prediction schemas | 1 | Tables deployed, no data, no consumers | Supporting |
| Signal normalization (0-1 scaling) | 0 | Does not exist — build | Core |
| Exponential decay weighting | 0 | Does not exist — build | Core |
| Composite scoring with confidence | 0 | Does not exist — build | Core |
| Insights UI (dashboard components) | 2 | Components exist, no data flowing | Generic |
Gaps
| Gap | Priority | Investment |
|---|---|---|
| Signal normalization | P0 | New function — per-signal-type normalization to 0-1 |
| Exponential decay | P0 | New function — recency weighting with configurable half-life |
| Composite scoring | P0 | New function — weighted sum with confidence calculation |
| Forecast baselines | P0 | Configuration — dream team sets targets per signal |
| Commissioning data access | P1 | Wire — parse L-levels from commissioning dashboard |
Build Ratio
~70% composition (algorithm framework, data access, constants pattern, Convex client), ~30% new code (normalization, decay, composite, signal collectors).
Questions
Which capability in the pipeline nowcast map is the hardest to replicate — and does that make it the moat?
- Which capabilities are table-stakes (must have) vs differentiators (competitive edge) — and is the build order aligned?
- If a new competitor entered with 10x engineering resources, which capabilities would they copy first and which would they build differently?
- Which capability gap, once closed, would unlock the most other capabilities?