Intelligence Functions
What's the difference between intelligence and agency?
Intelligence is what a system IS. Agency is what it CREATES. One is a feature. The other is an outcome. Every Mycelium capability that matters — alignment scoring, yield calculation, confidence gating, recommendation — lives here.
The Job
When a venture needs smart decisions, pattern detection, or recommendation logic, help it access pure, portable functions that produce outcomes — not features that require infrastructure.
| Trigger Event | Current Failure | Desired Progress |
|---|---|---|
| New venture needs decision logic | Rebuilds from scratch or copies brittle code | Imports from shared function library |
| Framework migration (Next → whatever) | Intelligence logic entangled with infrastructure | Functions survive migration untouched |
| Licensing opportunity | Value locked inside one app | Extractable, licensable, portable |
| Quality audit | Logic scattered across services, untestable in isolation | Pure functions, tested independently, zero infrastructure deps |
The Strategic Insight
| What Gets Commoditised | What Compounds |
|---|---|
| UI (generated, pattern-based) | Intelligence — real-world value creation |
| Plumbing (infrastructure, wiring) | Intelligence — hard to copy |
| Specs (templates, conventions) | Intelligence — zero infra, portable |
The 80% that's generated doesn't matter. The 20% that's human-written and outcome-driven IS the moat.
What Belongs Here
Pure functions with zero infrastructure dependencies:
- Alignment calculators — IDEAL vs REALITY gap scoring (Time + Mind)
- Yield formulas — shipped value / invested hours (Dreamineering Yield)
- Confidence scoring — how sure is the system about a decision (Workflow Engine)
- Deviation detection — patterns that signal drift from intention
- Recommendation engines — what to do next, based on what happened
- Commissioning algorithms — progressive maturity scoring (Commissioning)
The rule: If it has a database import, it doesn't belong here. If it survives being extracted to a standalone package, it does.
Feature / Function / Outcome
| # | Feature | Function | Outcome | State |
|---|---|---|---|---|
| 1 | Pure function library | Portable decision logic with no infra deps | Any venture imports and uses immediately | Gap |
| 2 | Alignment Calculator | Score IDEAL vs REALITY gap | Users see how close attention matches intention | Gap |
| 3 | Yield Formula | Calculate shipped artifacts / invested hours | Single number proves whether the work is working | Gap |
| 4 | Confidence Scorer | Quantify certainty of automated decisions | Humans intervene only when confidence is low | Gap |
| 5 | Deviation Detector | Flag when behaviour drifts from stated intention | Course correction before it compounds | Gap |
| 6 | Recommendation Engine | Suggest next action from patterns | Users act on insight, not guesswork | Gap |
Commissioning
| Component | Schema | API | UI | Tests | Status |
|---|---|---|---|---|---|
| Pure function library | Pending | Pending | N/A | Pending | 0% |
| Alignment Calculator | Pending | Pending | N/A | Pending | 0% |
| Yield Formula | Pending | Pending | N/A | Pending | 0% |
| Confidence Scorer | Pending | Pending | N/A | Pending | 0% |
| Deviation Detector | Pending | Pending | N/A | Pending | 0% |
Risks + Kill Signal
| Risk | Mitigation |
|---|---|
| Premature extraction — pulling functions out before they're stable | Extract only after a function is used by 2+ capabilities with tests passing |
| Over-abstraction — making it so generic it's useless | Each function must serve a concrete JTBD, not a theoretical pattern |
| Infra leaks in — database references sneak into "pure" functions | CI gate: intelligence layer builds with zero external dependencies |
Kill signal: If functions exist but no venture imports them (they rebuild instead), the abstraction is wrong. Rewrite or kill.
Mycelium Capability
Intelligence Functions are the foundation layer. Every other capability depends on them:
- Time + Mind uses Alignment Calculator and Yield Formula
- Workflow Engine uses Confidence Scorer and Escalation Router
- Commissioning uses progressive maturity algorithms
- ETL Data uses validation and quality scoring
- Agent Platform — agents execute these functions
Currently Growing In: All mushroom caps — this is platform infrastructure, not venture-specific.
Context
- Agent Platform — Agents that execute intelligence functions
- Products — What we build and why
- Standards — Where proven patterns graduate to
- Platform — Infrastructure that compounds
- Flow Engineering — The engineering methodology
- Phygital Mycelium — The capability catalogue