Demo

Run the full 10-step evaluation pipeline.

The MO§ES™ demo runs a complete operator evaluation on 50 synthetic operators across 5 AI platforms (ChatGPT, Claude, Codex, Copilot, Cursor). It scores, benchmarks, diagnoses, decomposes operator vs system effects, traces outcome lineage, and generates a full pilot readout. No real data. No API keys. No prompt text. Pure token telemetry.

Run it now Web walkthrough

The 10-step pipeline

What runs when you type one command.

Every step is real code from the platform. No mocks, no stubs. The demo loads synthetic data, runs the actual scoring engine, benchmark engine, diagnostic engine, decomposition analysis, and outcome correlation, then writes a markdown and PDF report.

LOAD

50 operators, 1,668 observations, 5 platforms, 50 lineages, 50 outcomes. Synthetic data with realistic distributions.

EVALUATE

Compute 5 canonical metrics per observation: leverage, yield, token SNR, construction, upsilon. The scoring engine runs on every observation.

BENCHMARK

Percentile-rank every operator against the cohort. 13 benchmark classes. Cross-operator, cross-system, cross-workflow comparisons.

DIAGNOSE

Pattern detection across 7 dimensions: capability, concentration, friction, dependency, instability, system differences, workflow fit. Every diagnosis is a hypothesis with evidence grade.

OPERATOR×SYSTEM

Two-way ANOVA-style decomposition. Partitions metric variance into operator effect, system effect, and interaction. Demo data shows operator effect dominates at 96-97%.

INTERVENE

Load intervention assignments. Each intervention has a target metric, a hypothesis, and an alternative explanation.

RE-EVALUATE

Re-score after intervention. Compute declared deltas. Verify whether the intervention moved the target metric.

OUTCOME LINEAGE

Trace the full lineage chain: observation to transformation to committed artifact to outcome. Correlate operating metrics with external quality scores and cycle times. ASSOCIATION, not causation.

REPORT

Generate a full pilot readout in markdown and PDF. Includes cohort stats, score distributions, metric distributions, divergence findings, intervention outcomes, and graphics.

VISUALIZE

9 graphics: measurement architecture, evaluation loop, connection architecture, operator-system matrix, transformation lineage, before/after intervention, capability topology, evidence ladder, sample customer report.

Run it

One command. No API keys. No real data.

The demo runs entirely on synthetic data. No external services are called. No prompt text is read. No real operator data is used. You need Python 3.10+ and the repo.

Option 1: One-liner (no clone needed)

curl -sL https://mos2es.org/demo/run.py | python3 -

Clones the repo to a temp directory, installs rich, runs all 10 steps, prints output. Requires Python 3.10+ and git.

Option 2: Clone and run

git clone https://github.com/SunrisesIllneverSee/b2bpilot.git
cd b2bpilot/_01_platform
pip install rich
python3 -m src.cli.main demo full

That's it. The demo runs all 10 steps and writes outputs to demo_data/graphics/:

demo_full_pilot_readout.md — full markdown pilot report
g09_sample_customer_report.pdf — PDF version of the report
9 diagram files (Mermaid + text + HTML)

Option 3: Run individual commands

# Score a single operator
python3 -m src.cli.main score operator op_001

# Compare operators across the cohort
python3 -m src.cli.main compare cohort

# Run diagnostics on an operator
python3 -m src.cli.main diagnose operator op_001

# Operator×System decomposition
python3 -m src.cli.main compare operator-system

# Lineage chain for an operator
python3 -m src.cli.main lineage show op_046

# Outcome correlation across the cohort
python3 -m src.cli.main lineage outcomes

# Org AI topology (team-level analysis)
python3 -m src.cli.main compare topology

# Operator similarity search
python3 -m src.cli.main compare similarity op_001

# Export the executive dashboard as HTML
python3 -m src.cli.main export dashboard --output dashboard.html

Option 4: Run the test suite

cd b2bpilot/_01_platform
python3 -m pytest tests/ -q
# 527 tests pass

Option 5: Run the MCP server

cd b2bpilot/_01_platform
pip install mcp
python3 -m src.mcp_server.server

The MCP server exposes 25 tools that Claude, ChatGPT, Cursor, and other AI agents can call natively. See the developer docs for the full tool list and API spec.

Prefer a visual walkthrough?

The interactive web demo.

If you want to see what the product looks like without running code, the web walkthrough at enterprise.mos2es.org walks through each stage of the evaluation pipeline with visual explanations of the canonical primitives, metrics, diagnostics, workflow fit, and intervention testing.

Open the web walkthrough

Outputs

What the demo produces.

Pilot readout (markdown + PDF)

A full sample customer report with cohort statistics, score distributions, metric distributions, divergence findings, intervention outcomes, and the evidence ladder. This is the same format a real pilot delivers.

9 architecture diagrams

Mermaid + text diagrams covering measurement architecture, evaluation loop, connection architecture, operator-system matrix, transformation lineage, before/after intervention, capability topology, and evidence ladder.

Executive dashboard (HTML)

Self-contained HTML dashboard with inline SVG charts. Shows composite score distribution, cohort-level metric summaries, and operator rankings. Generated by enterprise export dashboard.

Console output

Every step prints structured output to the console. Add --json to any command for machine-readable JSON output suitable for piping into other tools or agents.

The synthetic dataset

50 operators. 5 platforms. 1,668 observations.

The demo data is fully synthetic. No real operators, no real prompt text, no real API calls. The dataset is designed to exercise every analysis path in the platform:

50 operators with pseudonymous IDs (op_001 through op_050)
5 AI platforms: ChatGPT, Claude, Codex, Copilot, Cursor
1,668 observations with INPUT, OUTPUT, CACHE READ, CACHE WRITE token counts
50 lineages linking observations to transformations to artifacts to outcomes
50 outcomes with external quality scores and cycle times
5 teams for org topology analysis
7 workflow stages for workflow fit analysis
Intervention assignments with pre/post measurement

Synthetic markers are preserved across all import/export operations. The demo data is deterministic — running the demo twice produces identical results.