DN Agent Team Cost and Capacity Planner v1.0
Research package for Decentralised News, Batch 2 article 73.

START HERE
Open agent-team-cost-and-capacity-planner.html in a current browser. It runs
locally without an internet connection, external libraries or network requests.
Edit weekly incoming work items, profile, owner review budget, fixed monthly
software cost and owner hourly value. The five-role mix and profile assumptions
remain fixed for this version. Results update as inputs change.

WHAT THIS PACKAGE ESTABLISHES
This is an authored synthetic planning model with zero observed deployments.
All workload shares, rerun shares, acceptance rates, review times and CU costs
are fictional assumptions. CU denotes fictional cost units, not vendor prices.
The 27 cases are combinations of three profiles, three weekly workloads and
three owner-time budgets. They are not 27 experiments or measured businesses.
Work items differ across roles and cannot be treated as interchangeable clients.
Owner-time value is noncash. No revenue, profit, savings or payback is modeled.

FILES
assumptions.json: the numerical inputs, definitions and technical sources.
roles.csv: five role-level input records.
scenario-results.csv: original generator output containing all 27 scenarios.
scenarios.json: the same scenarios as machine-readable records.
summary.json: the Planning base case, scenario counts and capacity ceilings.
MODEL-METHODOLOGY.txt: formulas, units, scope and limitations.
build_model.py: reproducible Python standard-library generator.
agent-team-cost-and-capacity-planner.html: editable standalone offline planner.
model-review-report.json: independent assistant arithmetic and disclosure review.
PARITY-VERIFICATION.txt: separate Python/HTML and reproducibility check record.
SHA256SUMS: checksums for every other file in this package.

REPRODUCE THE SCENARIO EXPORTS
Requires Python 3 and its standard library; no packages or credentials needed.
After extracting this directory, run:

    python build_model.py --output reproduced

The reproduced directory receives assumptions.json, roles.csv,
scenario-results.csv, scenarios.json, MODEL-METHODOLOGY.txt and summary.json.
The HTML planner and review records are included artifacts, not regenerated by
that command. Read MODEL-METHODOLOGY.txt before interpreting the calculations.

VERIFY THE PACKAGE FILES
From the extracted package directory, where sha256sum is available, run:

    sha256sum -c SHA256SUMS

SHA256SUMS covers the eleven other files; it does not attempt to hash itself.
The separately distributed public CSV named
'dn-agent-team-cost-and-capacity-scenarios-v1.0.csv' is byte-identical to the
original scenario-results.csv retained in this package.

BASE CASE AND LIMITS
Planning, 100 incoming items/week, 15 owner hours/week: total owner time is
15.2166667 hours, exceeding the budget by 13 minutes. The capacity floor is 98
incoming items/week under the fixed mix. The 90 planned accepted items are not
realized throughput when the owner-time budget is exceeded. Monthly software
cost is 408.5416667 CU; full modeled cost is 3046.0972222 CU including the
noncash owner-time valuation. The maximum-two-attempt software scenario is a
conditional model amount, not an enforced bill cap.

The independent review establishes internal consistency only. Neither that
review nor the offline planner establishes attainable business throughput,
reviewer accuracy, autonomous permissions, billing controls or real-world ROI.
No WordPress publishing data, credentials or upload receipts are included.
