Essential Algorithms
What is THE algorithm that makes your business work?
Strip away the product, the brand, the team. What remains is a routing function. Every business takes an intent, finds a path, delivers through infrastructure, settles value, and feeds the result back. The business that understands its own algorithm controls its own leverage. The one that doesn't is running someone else's.
The Pattern
INTENT → ROUTE → INFRASTRUCTURE → SETTLE → FEEDBACK
| Flow | Intent | Route | Infrastructure | Settle | Feedback |
|---|---|---|---|---|---|
| Telecom | Voice call | Carrier switches | Backbone network | CDR billing | Quality metrics |
| Crypto | Token swap | Solvers/Bridges | Blockchain | On-chain settlement | Slippage data |
| DePIN | Compute/data | Edge AI | Device mesh | Oracles + chain | Network health |
| SaaS | Job to do | Matching engine | Platform | Subscription | Usage signals |
| Commerce | Purchase | Creator storefront | Payment rails | Transaction | Repeat rate |
| Community | Belonging | Contribution graph | Third spaces | Trust proof | Retention |
Same five steps. Different matter. Whoever controls the routing intelligence captures the arbitrage.
Three Layers
The algorithm exists at three zoom levels. Theory defines the rules. Code executes them. Commerce wraps them in a shop front.
| Layer | What Lives Here | Example |
|---|---|---|
| Theory | Business principles as algorithms — distribution, leverage, moats, network effects | The XV are fifteen algorithms dressed as principles |
| Code | Pure functions with tunable constants — scoring, routing, decision, optimization | calculateSPCL(), exploreExploit(), optimalStopping() |
| Commerce | Venture shop fronts that present algorithm output as customer value | Stackmates, Prettymint, Howzus |
Theory without code is opinion. Code without commerce is a side project. Commerce without theory is guesswork. The wiring between layers is the business.
Algorithm Registry
Coded algorithms in @stackmates/agency. Each has tunable constants and a domain.
| Algorithm | Domain | Category | Function | Venture Surface |
|---|---|---|---|---|
| SPCL Scoring | Marketing | Scoring | Content specificity, personality, conversation, locality | Dreamineering, Berley Trails |
| Explore-Exploit | Strategy | Decision | Multi-armed bandit with UCB and epsilon decay | All (resource allocation) |
| Optimal Stopping | Strategy | Decision | Secretary problem — 37% look, then commit | Hiring, PRD prioritization |
| Sales Forecasting | Sales | Prediction | Deal probability with recency weighting | Stackmates |
| DePIN Score | Crypto | Scoring | Network health × token economics × team × tech | Stackmates (crypto vertical) |
| AI Tuning Loop | AI | Optimization | Gradient-based parameter tuning with momentum | All (agent improvement) |
| Eval Runner | AI Quality | Evaluation | LLM-as-Judge for response quality | All (quality assurance) |
| Industry Solar | Analysis | Classification | Confidence-scored industry matching | Deep research |
| RFP Type Detection | Sales | Routing | Document classification and routing | Stackmates |
| Compound Rate Tracking | RFP | Analysis | Library growth rate measurement | Stackmates |
| Meta | Force | VVFL as a force function — the meta-algorithm | Dormant |
Tunable Constants
Every algorithm has constants that encode assumptions. When assumptions change, constants change. When constants change, behavior changes.
| Algorithm | Key Constants | Current Values |
|---|---|---|
| SPCL | Weight per dimension | 0.25 each (equal) |
| Explore-Exploit | Initial epsilon, decay rate, UCB constant | 0.3, 0.995, 2.0 |
| Optimal Stopping | Look ratio (1/e), min candidates | 0.368, 3 |
| Sales Forecasting | Recent vs historical weight | 0.7 / 0.3 |
| DePIN Score | Health, tokenomics, team, tech weights | 0.3, 0.25, 0.2, 0.25 |
| AI Tuning | Learning rate, momentum, min improvement | 0.01, 0.9, 0.001 |
Constants are the setpoints. The algorithm is the PID controller. Results are the process variable. When outputs diverge from expectations, tune the constants — don't rewrite the algorithm.
Venture Wiring
Each venture IS an algorithm wearing a shop front.
| Venture | Essential Algorithm | What Gets Routed | MEV Source |
|---|---|---|---|
| Stackmates | Matching + forecasting | Teams to tools to workflows | Platform lock-in through workflow intelligence |
| Prettymint | Loyalty loop scoring | Creators to buyers to repeat members | Repeat purchase compounding |
| Howzus | Contribution graph routing | Builders to projects to proof | Trust accumulation from verified contribution |
| Dreamineering | SPCL content scoring | Ideas to articles to conviction | Attention routed to high-signal content |
| Better Practice | AI tuning + eval | Practice reps to feedback to improvement | Learning rate through validated reps |
| Berley Trails | Explore-exploit | Curiosity to discovery to content | Surface area for luck |
| Touch For Fun | Trust delta measurement | Sessions to rituals to belonging | Protocol quality from measured trust |
The mycelium shares the 95% — identity, CRM, workflows, analytics. Each venture's 5% is its essential algorithm.
The Meta-Algorithm
The VVFL is the algorithm that improves all other algorithms.
QUESTIONS → PRINCIPLES → PROTOCOLS → STANDARDS → PLATFORM → BETTER QUESTIONS
It's coded as vfl-force-potential but dormant — because it needs the other algorithms running and producing data before it can measure force. When it wakes up, it becomes the governor: which algorithm is producing the most value per unit energy? Route more resources there. Which is stalling? Tune constants or kill it.
The escape velocity game is what happens when the meta-algorithm runs long enough. Seven stages, three phase transitions. The algorithm transforms what it optimizes at each stage — from product to platform to standard to infrastructure. Same code, different constants, different routing table.
The XV as Algorithms
The Business XV are fifteen algorithms that most businesses run unconsciously. Making them explicit is the first step to tuning them.
| Principle | Algorithm | Input | Output |
|---|---|---|---|
| Distribution | Channel routing | Product + audience | Reach |
| Leverage | Force multiplication | Capital + labor + code + media | Asymmetric returns |
| Network Effects | Value scaling | n users | n² connections |
| Moat | Competitive isolation | Advantages | Switching costs |
| Snowball | Compound accumulation | Platform × process × people | Accelerating returns |
| Info Arbitrage | Knowledge routing | Asymmetric information | Trading profit |
| Unit Economics | Per-unit math | CAC + LTV + margin | Viability signal |
Each principle page on this site is a spec for an algorithm most businesses run by instinct. Running by instinct means you can't tune the constants.
Context
- Telco MEV Algorithm — The proof: a real routing algorithm that generated real arbitrage
- Business XV — Fifteen algorithms dressed as principles
- Software Algorithms — CS reference catalog
- Decision Algorithms — Human heuristic algorithms
- Pricing Algorithm — Demand-capacity matching
- VVFL — The meta-algorithm
- Escape Velocity — What happens when algorithms compound
- Information Arbitrage — Whoever controls routing captures value
- Dreamineering Ventures — Shop fronts for coded algorithms
Questions
What is YOUR essential algorithm — the routing function that, if removed, means no business exists?
- Which constants in your algorithm are you tuning by instinct instead of measurement?
- Where does your routing intelligence live — in people's heads, in code, or nowhere?
- If your competitor understood your algorithm better than you do, what would they exploit?
- At what stage of escape velocity does your algorithm need to change what it optimizes?