---
id: e6-compute-economics-package
title: Compute Economics Package
module: GROW-S7
module_slug: grow-s7-compute-economics
cluster: Execution
type: template
version: v0.1.0
status: Gate-reviewed
tier: membership
contract_role: ""
canonical_url: "https://grow.goodcombinator.ai/library/registry/e6-compute-economics-package"
download_url: "https://grow.goodcombinator.ai/library/registry/e6-compute-economics-package.md"
license: CC-BY-4.0 (proposed — owner confirmation required)
source: GROW by Good Combinator
retrieved_at: 2026-05-29
---

# Compute Economics Package

The Compute Economics Package is the practitioner's on-ramp for GROW-S7. It bundles six fillable scaffolds — one for each of the five substantive Course 6 artifacts plus an optimization roadmap — so a builder can stand up a complete compute-economics discipline for a new agent project by completing a single package rather than opening each artifact independently. Each scaffold section is a minimal, project-specific instantiation of the full spec; builders should cross-reference the source artifact for the full rules, worked examples, and change-control requirements. This package is the deployment artifact; the individual artifacts in this directory are the authoritative specifications.

---

## How to Use This Package

1. **Copy this file** to a project-specific location and rename it `<agent-id>-compute-economics-package.md`.
2. **Fill all six scaffolds in order.** Each scaffold has a declared upstream dependency; complete them top-to-bottom. Do not skip the cost model — every downstream scaffold derives from it.
3. **Reference source artifacts** for edge cases, rules, and worked examples. The canonical ids are cross-referenced in each scaffold header.
4. **Update on each billing cycle.** The cost-model and unit-economics rows are living data; the routing policy and latency budgets should be re-reviewed on each major model or infrastructure change.
5. **Verify C12 compliance** after filling scaffold 6. Before any commercialization work cites this package, confirm that every `unit_cost_ref` in `c10-revenue-model-design` resolves to a `viable: true` row in scaffold 6.

---

## Scaffold 1 — Cost Model

*Source artifact: `e6-cost-model` (spec)*

```yaml
# PROJECT-SPECIFIC COST MODEL
agent_id: <kebab-case>
billing_period: monthly

# Fill each category; mark "N/A" if not applicable — do not omit
model_api:
  premium_cost_per_call_usd: <number>
  standard_cost_per_call_usd: <number>
  cheap_cost_per_call_usd: <number>
  estimated_calls_per_month:
    premium: <number>
    standard: <number>
    cheap: <number>
    deterministic: <number>

infrastructure_monthly_usd: <number>
storage_monthly_usd: <number>
retrieval_monthly_usd: <number>
tooling_monthly_usd: <number>
human_review_monthly_usd: <number>  # REQUIRED if any HITL gates exist
observability_monthly_usd: <number>

total_monthly_estimate_usd: <derived>

budget_ceilings:
  hard_monthly_usd: <number>
  soft_monthly_usd: <number — must be < hard>
  hard_per_run_usd: <number>
  hard_token_per_call: <integer>

cost_per_outcome:
  cost_per_task_usd: <derived>
  cost_per_user_usd: <derived>
```

---

## Scaffold 2 — Quality-Cost Matrix

*Source artifact: `e6-quality-cost-matrix` (matrix)*

List every workflow step that involves a model call or a decision branch. Steps not listed here default to `minimum_tier: premium` (safe default; flag for review).

```yaml
# PROJECT-SPECIFIC QUALITY-COST MATRIX
steps:
  - step_id: <kebab-case>
    quality_requirement: <high | medium | low>
    max_severity_present: <critical | high | medium | low | info>
    minimum_tier: <premium | standard | cheap | deterministic>
    demotion_allowed: <true | false>  # false if max_severity_present is critical or high
    demotion_waiver_ref: <artifact-id or "none">
    rationale: <one sentence>

  # Add rows for each step...

# Tier distribution summary (fill after listing all steps)
tier_summary:
  premium_steps: <count>
  standard_steps: <count>
  cheap_steps: <count>
  deterministic_steps: <count>
  estimated_premium_spend_pct: <number>
```

---

## Scaffold 3 — Routing Policy

*Source artifact: `e6-routing-policy` (policy)*

```yaml
# PROJECT-SPECIFIC ROUTING POLICY
routing_policy_id: <agent-id>-routing-policy
agent_id: <matches scaffold 1>
cost_model_ref: e6-cost-model          # or project-specific instance
quality_matrix_ref: e6-quality-cost-matrix
latency_budget_ref: e6-latency-budget-spec

cache:
  enabled: <true | false>
  similarity_threshold: <0.90–0.99>
  max_cache_age_seconds: <integer>
  excluded_step_ids: [<list step-ids that must never be cached>]

fallback_chain:
  premium_fallback: <standard | human-review | abort>
  standard_fallback: <cheap | human-review | abort>
  cheap_fallback: <deterministic | abort>

rate_limit_policy:
  max_retries: <integer — max 3 per e6-waste-reduction-playbook W1 cap>
  backoff_seconds: <integer>
  max_planning_cycles_per_run: <integer — max 10 per e6-waste-reduction-playbook W3 cap>

# C7 emission reminder: append {model_tier, routing_rationale, est_cost_usd, est_latency_ms,
# fallback_tier_if_any} to s2-audit-trail-schema tool_calls[] on every call.

# C8 emission reminder: emit C2 six-field HITL event to s1-hitl-review-policy when
# hard_monthly_usd or hard_per_run_usd is crossed. decision_origin: escalation.
# Tier never alone flips a safety or quality verdict.
```

---

## Scaffold 4 — Latency Budget

*Source artifact: `e6-latency-budget-spec` (spec)*

```yaml
# PROJECT-SPECIFIC LATENCY BUDGET
agent_id: <matches scaffold 1>

task_classes:
  - task_class_id: <kebab-case>
    interaction_type: <interactive | async | batch>
    ceiling_ms: <integer>
    p95_target_ms: <integer>
    p99_ceiling_ms: <integer>
    model_tier_ceiling: <premium | standard | cheap | deterministic>
    bottleneck_layer: <model-inference | retrieval | tool-call | orchestration | network | human-review>
    bottleneck_mitigation: <one sentence>
    cache_eligible: <true | false>
    batch_eligible: <true | false>

  # Add rows for each task class...

latency_alerts:
  p99_breach_action: hard-gate
  escalation_target_ref: s1-threshold-escalation-spec
```

---

## Scaffold 5 — Waste Reduction Register

*Source artifact: `e6-waste-reduction-playbook` (playbook)*

Track active waste patterns and their remediation status for this project.

```yaml
# PROJECT-SPECIFIC WASTE REDUCTION REGISTER
agent_id: <matches scaffold 1>
last_audit_date: <ISO-8601>

active_waste_patterns:
  - pattern_id: <W1 | W2 | W3 | W4 | W5 | W6>
    description: <one sentence describing this project's specific instance>
    detected_date: <ISO-8601>
    estimated_monthly_waste_usd: <number>
    remediation_action: <one sentence>
    remediation_owner: <role>
    target_completion: <ISO-8601>
    status: <open | in-progress | resolved>
    resolution_note: <required when status=resolved>

  # Add rows for each active pattern...

waste_audit_cadence: monthly
next_audit_date: <ISO-8601>
```

---

## Scaffold 6 — Unit Economics Worksheet

*Source artifact: `e6-unit-economics-worksheet` (template)*

This scaffold is the C12 cost source. Fill one row per offer before any commercialization work begins. `c10-revenue-model-design` must cite `unit_cost_ref` pointing to a populated row here.

```yaml
# PROJECT-SPECIFIC UNIT ECONOMICS WORKSHEET
agent_id: <matches scaffold 1>

offers:
  - offer_id: <kebab-case — this is the unit_cost_ref used in c10-revenue-model-design>
    offer_name: <short name>
    offer_type: <task | workflow | user-seat | outcome>

    # All cost inputs from scaffold 1 allocated per unit
    model_api_cost_per_unit_usd: <number>
    compute_cost_per_unit_usd: <number>
    storage_retrieval_cost_per_unit_usd: <number>
    tooling_cost_per_unit_usd: <number>
    human_review_cost_per_unit_usd: <number — MUST NOT be zero if HITL gates exist>
    egress_observability_cost_per_unit_usd: <number>
    total_cogs_per_unit_usd: <derived>

    revenue_per_unit_usd: <number>
    revenue_model: <subscription-allocated | usage-fee | outcome-fee | internal-cost-avoidance>
    units_per_month: <number>

    gross_margin_pct: <derived>
    gross_margin_usd_per_unit: <derived>
    monthly_gross_profit_usd: <derived>

    fixed_cost_allocation_usd: <number>
    break_even_units_per_month: <derived>

    automation_increases_cost: <true | false>
    manual_baseline_cost_per_unit_usd: <required if automation_increases_cost=true>
    viable: <true | false>
    viability_note: <required if viable=false>
    min_margin_pct: <number — must match c10-revenue-model-design's min_margin_pct for this offer>
    last_updated: <ISO-8601>

  # Add rows for each offer...

# C12 ANTI-VAPORWARE CHECK:
# Before any pricing hypothesis in c10-revenue-model-design cites this worksheet,
# confirm: (1) every offer_id cited resolves to a row above, (2) viable=true,
# (3) gross_margin_pct >= min_margin_pct. Flag any viable=false row to the
# commercialization team before they build a pricing hypothesis on it.
```

---

## Optimization Roadmap

The optimization roadmap captures planned compute-economics improvements beyond immediate waste remediation — model tier changes, infrastructure migrations, cache strategy upgrades, or pricing model pivots. Unlike the waste reduction register (which tracks active waste), the roadmap tracks forward-looking improvements with investment and expected return.

```yaml
# PROJECT-SPECIFIC OPTIMIZATION ROADMAP
agent_id: <matches scaffold 1>

optimization_initiatives:
  - initiative_id: <kebab-case>
    description: <one sentence>
    category: <tier-optimization | cache-strategy | infra-right-sizing | hitl-reduction | pricing-model-change>
    expected_monthly_savings_usd: <number>
    investment_required_usd: <number>
    payback_period_months: <derived>
    depends_on: [<list of prerequisite initiatives or artifacts>]
    owner: <role>
    target_date: <ISO-8601>
    status: <planned | in-progress | completed | deferred>

  # Add rows for each initiative...

roadmap_review_cadence: quarterly
next_review_date: <ISO-8601>
```

---

## Package Validation Checklist

Before marking this package `status: reviewed`, confirm all of the following:

- [ ] Scaffold 1: `hard_monthly_usd` is set and `soft_monthly_usd < hard_monthly_usd`
- [ ] Scaffold 1: `human_review_monthly_usd` is populated (not zero) if any HITL gates exist
- [ ] Scaffold 2: every workflow step with a model call has a matrix row
- [ ] Scaffold 2: no `critical` or `high` severity step has `demotion_allowed: true` without a `demotion_waiver_ref`
- [ ] Scaffold 3: `max_retries ≤ 3` and `max_planning_cycles_per_run ≤ 10`
- [ ] Scaffold 3: C7 and C8 emission reminders are acknowledged in code implementation
- [ ] Scaffold 4: every task class has a `p99_ceiling_ms` and an escalation target
- [ ] Scaffold 5: waste audit date is set; any active W3 (runaway loop) pattern is `status: in-progress` or `resolved`
- [ ] Scaffold 6: every offer has a `viable` flag and `viable=false` rows have a `viability_note`
- [ ] Scaffold 6: `gross_margin_pct ≥ min_margin_pct` for all `viable: true` rows
- [ ] Optimization roadmap: at least one initiative is planned if any waste patterns are open