Read-only dashboards with energy, governance, stress scenarios and preserved diagrams. No wallet or installation required; all values come from recorded simulations.
Inspect task/artifact hashes, independently recompute the energy ledger, then require independent substantive review and a separately authorized signer.
# Kardashev Business 3 — Foundation
This is the ASCII-safe entry point for the **Foundation simulation engine**. AGI Jobs coordinates authorized, lawful screen-based work with specialized agents, reviewable evidence, independent verification and separate settlement authority.
## Start here
- [Open the shared website](https://montrealai.github.io/AGIJobsv0/experiments/kardashev-business/) after deployment.
- [Read the complete workbench guide](../kardashev_ii_omega_grade_alpha_agi_business_3/workbench/README.md): ten job templates, local artifact checks, approved OpenClaw dispatch, capacity assumptions and qualification limits.
- From the repository root, run `npm run demo:kardashev-business:serve` for the local explorer. No dependency installation is needed for that command.
## Run this simulation
Use Python 3.12 from the repository root:
```sh
python -m demo.kardashev_ii_omega_grade_alpha_agi_business_3 --max-cycles 3 --no-resume
```
This bounded command produces synthetic orchestration state. It does not invoke OpenAI, operate a real desktop or submit blockchain transactions. `npm run demo:kardashev-business:smoke` checks all five engines using isolated temporary state.
The original implementations and flowcharts remain available through the [demo catalog](../README.md). The shared website adds a practical work-order and evidence path; simulation output is not production qualification.
{"revision":"5b4cebb309a83a7a6749d8911d8bf96a1921e042","sources":[{"file":"demo/kardashev_ii_omega_grade_alpha_agi_business_3/README.md","content":"# Kardashev Business 3 — Foundation\n\nThis is the ASCII-safe entry point for the **Foundation simulation engine**. AGI Jobs coordinates authorized, lawful screen-based work with specialized agents, reviewable evidence, independent verification and separate settlement authority.\n\n## Start here\n\n- [Open the shared website](https://montrealai.github.io/AGIJobsv0/experiments/kardashev-business/) after deployment.\n- [Read the complete workbench guide](../kardashev_ii_omega_grade_alpha_agi_business_3/workbench/README.md): ten job templates, local artifact checks, approved OpenClaw dispatch, capacity assumptions and qualification limits.\n- From the repository root, run `npm run demo:kardashev-business:serve` for the local explorer. No dependency installation is needed for that command.\n\n## Run this simulation\n\nUse Python 3.12 from the repository root:\n\n```sh\npython -m demo.kardashev_ii_omega_grade_alpha_agi_business_3 --max-cycles 3 --no-resume\n```\n\nThis bounded command produces synthetic orchestration state. It does not invoke OpenAI, operate a real desktop or submit blockchain transactions. `npm run demo:kardashev-business:smoke` checks all five engines using isolated temporary state.\n\nThe original implementations and flowcharts remain available through the [demo catalog](../README.md). The shared website adds a practical work-order and evidence path; simulation output is not production qualification.\n","format":"text","sha256":"c268928d2ac1745157432e45b89396495629dc1864d8d630e5e04390646ffb72","bytes":1421,"download":"/AGIJobsv0/examples/c268928d2ac17451-README.md","source":"https://github.com/MontrealAI/AGIJobsv0/blob/5b4cebb309a83a7a6749d8911d8bf96a1921e042/demo/kardashev_ii_omega_grade_alpha_agi_business_3/README.md"},{"file":"demo/kardashev_ii_omega_grade_alpha_agi_business_3/workbench/catalog.mjs","content":"export const families = [\n {\n id: 'foundation',\n title: 'Foundation',\n focus: 'Resource-ledger and recursive-job foundations.',\n module: 'demo.kardashev_ii_omega_grade_alpha_agi_business_3',\n args: '--max-cycles 3 --no-resume',\n },\n {\n id: 'business',\n title: 'Business 3',\n focus:\n 'Worker coordination, validation windows, integrity and owner controls.',\n module: 'demo.kardashev_ii_omega_grade_alpha_agi_business_3_demo',\n args: '--cycles 3 --no-resume',\n },\n {\n id: 'omega',\n title: 'Omega',\n focus: 'An operator mission plan with recursive job graphs.',\n module: 'demo.kardashev_ii_omega_grade_alpha_agi_business_3_demo_omega',\n args: '--cycles 3 --duration 3',\n },\n {\n id: 'supreme',\n title: 'Supreme',\n focus: 'Message-bus orchestration, checkpoints and planetary telemetry.',\n module: 'demo.kardashev_ii_omega_grade_alpha_agi_business_3_demo_supreme',\n args: '--cycles 3 --no-resume --validator_commit_delay_seconds 1 --validator_reveal_delay_seconds 1',\n },\n {\n id: 'ultra',\n title: 'Ultra',\n focus: 'Mission archives and bounded orchestration with owner commands.',\n module: 'demo.kardashev_ii_omega_grade_alpha_agi_business_3_demo_ultra',\n args: 'launch --cycles 3 --no-sim',\n },\n];\nexport const templates = [\n {\n id: 'energy',\n title: 'Energy & scientific analysis',\n role: 'Numerical analyst',\n goal: 'Audit a synthetic energy ledger and identify every deficit.',\n input: '',\n artifact: 'analysis.json',\n criteria: [\n 'Preserve each source row exactly once and compute netKwh as generated minus consumed.',\n 'Report the exact total netKwh and all deficit IDs in source order.',\n 'Include schemaVersion: 1, kind: energy-balance, sourceSha256, rows: [{id, netKwh}], summary: {netKwh, deficits}, productionApproved: false and settlementApproved: false. Use decimal strings for all energy amounts.',\n ],\n },\n {\n id: 'performance',\n title: 'Performance optimization',\n role: 'Performance engineer',\n goal: 'Design and benchmark a bounded optimization for a synthetic deduplication workload.',\n input:\n 'Synthetic input generator: integers i % 997 for i from 0 through 99999. Preserve first-occurrence order. Compare a quadratic baseline with a set-based approach. Record runtime, versions and benchmark method; do not invent benchmark results.',\n artifact: 'benchmark.md',\n criteria: [\n 'Demonstrate output equivalence including empty input and duplicates.',\n 'Include runnable code, environment and measured timing or explicitly mark the benchmark unexecuted.',\n ],\n },\n {\n id: 'feature',\n title: 'Open-source features',\n role: 'Software engineer',\n goal: 'Implement an accessible CSV preview component as reviewable source text.',\n input:\n 'Synthetic CSV: name,units\\nSolar,12\\nStorage,8. Support quoted commas, empty cells and keyboard navigation. No external data or deployment.',\n artifact: 'implementation.md',\n criteria: [\n 'Include source code, build instructions and tests for quoted and malformed CSV.',\n 'Explain keyboard behavior and label any checks that were not executed.',\n ],\n },\n {\n id: 'sdk',\n title: 'API & SDK tooling',\n role: 'Integration engineer',\n goal: 'Produce a typed client specification for a synthetic job-status API.',\n input:\n 'GET /jobs/{id} returns {id:string,status:posted|working|review|complete}. Read-only requests only. Include timeout, cancellation and bounded GET retries; no transaction methods.',\n artifact: 'sdk.md',\n criteria: [\n 'Supply runnable client code and fixture tests for each status and malformed replies.',\n 'Document error handling, cancellation and limits.',\n ],\n },\n {\n id: 'tests',\n title: 'Automated testing',\n role: 'Quality engineer',\n goal: 'Write a reproducible test package for a synthetic invoice calculator.',\n input:\n 'totalCents = quantity * unitCents + shippingCents. All inputs are nonnegative integers. Reject values above 1000000. Accept zero. No tax, payment or live invoice system.',\n artifact: 'tests.md',\n criteria: [\n 'Cover boundaries, zero, invalid inputs and integer totals.',\n 'Include runnable commands and distinguish executed tests from proposed tests.',\n ],\n },\n {\n id: 'evals',\n title: 'AI evaluation packages',\n role: 'Evaluation designer',\n goal: 'Create a synthetic evaluation set for evidence-grounded summarization.',\n input:\n 'Source: Array A generated 120 kWh and consumed 90 kWh. Array B generated 50 kWh and consumed 70 kWh. Nothing is known about real-world energy production.',\n artifact: 'evaluation.json',\n criteria: [\n 'Include at least five cases, expected facts and a reproducible scoring rubric.',\n 'Include unsupported-claim and omission cases; do not equate this small eval with general intelligence.',\n ],\n },\n {\n id: 'dashboard',\n title: 'Data dashboards',\n role: 'Data visualization engineer',\n goal: 'Build a small accessible dashboard from synthetic energy data.',\n input:\n 'Rows: north generated=12500 consumed=8700; east generated=8400 consumed=9100; west generated=10200 consumed=6400. Units kWh. Preserve negative balances.',\n artifact: 'dashboard.md',\n criteria: [\n 'Include self-contained source text and a table fallback.',\n 'Show east as a deficit; disclose units and synthetic provenance.',\n ],\n },\n {\n id: 'research',\n title: 'Product intelligence',\n role: 'Research analyst',\n goal: 'Compare synthetic suppliers against an explicit delivery constraint.',\n input:\n 'Quantity 40, max delivery 7 days. Boreal: 12 USDC/unit + 50 shipping, 5 days. Laurentian: 10.5/unit + 40 shipping, 12 days. Stellar: 11.5/unit + 30 shipping, 7 days. No purchases.',\n artifact: 'comparison.md',\n criteria: [\n 'Show exact total costs and eligibility for all suppliers.',\n 'Recommend the lowest-cost eligible supplier and disclose that all quotes are synthetic.',\n ],\n },\n {\n id: 'documents',\n title: 'Documents & executable guides',\n role: 'Technical writer',\n goal: 'Prepare an operator guide for an authorized, bounded computer-work task.',\n input:\n 'Workflow: task definition, permission check, specialist execution, automated checking, independent review, separate settlement. Only public/licensed/synthetic inputs. No secrets or private files.',\n artifact: 'guide.md',\n criteria: [\n 'Include prerequisites, a reproducible local example, expected outputs and stop/recovery steps.',\n 'Separate artifact checking, independent review and signer approval.',\n ],\n },\n {\n id: 'reproduction',\n title: 'Numerical reproduction',\n role: 'Scientific computing specialist',\n goal: 'Reproduce a synthetic energy balance calculation with exact arithmetic.',\n input:\n 'Generated kWh: [12500,8400,10200]. Consumed kWh: [8700,9100,6400]. Compute each difference, sum, and deficit indices. Explain assumptions and provide executable code.',\n artifact: 'reproduction.md',\n criteria: [\n 'Report differences [3800,-700,3800] and total 6900 kWh.',\n 'Provide runnable verification and separate simulated quantities from measured physical output.',\n ],\n },\n];\nexport const energySource = {\n kind: 'energy-balance',\n goal: 'Reconcile synthetic array generation and consumption in kWh.',\n rows: [\n { id: 'north', generatedKwh: '12500', consumedKwh: '8700' },\n { id: 'east', generatedKwh: '8400', consumedKwh: '9100' },\n { id: 'west', generatedKwh: '10200', consumedKwh: '6400' },\n ],\n};\n","format":"text","sha256":"c5bfd5f3357318a1ff3f62909cc4ec399742992dee3bb721a9c8cf0f00a22a3a","bytes":7689,"download":"/AGIJobsv0/examples/c5bfd5f3357318a1-catalog.mjs","source":"https://github.com/MontrealAI/AGIJobsv0/blob/5b4cebb309a83a7a6749d8911d8bf96a1921e042/demo/kardashev_ii_omega_grade_alpha_agi_business_3/workbench/catalog.mjs"},{"file":"demo/kardashev_ii_omega_grade_alpha_agi_business_3/workbench/README.md","content":"# Kardashev Business 3 — machine labor workbench\n\nAGI Jobs is designed as a scalable machine labor layer for authorized, lawful screen-based work: specialized agents execute tasks, produce reviewable evidence, and support independent verification and settlement. This shared workbench connects all five Business 3 demos to that workflow while preserving their original simulation engines and diagrams.\n\n**Start here:** [Open the website](https://montrealai.github.io/AGIJobsv0/experiments/kardashev-business/) after this change is deployed, or run the local command below. The website is an offline planner and artifact checker. It does not connect a wallet, dispatch a worker, or operate physical systems.\n\n## Quick start\n\nFrom the repository root with the repository-pinned Node version:\n\n```sh\nnpm run demo:kardashev-business:serve\n```\n\nOpen `http://127.0.0.1:4188`. Select one of ten workflows, download its work order, generate an energy-ledger example, or test a wrong answer. The local explorer and deterministic example require only Node's standard library. Stop the server with Ctrl+C. `PORT=4190` changes the viewer port; the synthetic tasks still carry the documented default origin and need no network access.\n\nTo produce and check actual local files:\n\n```sh\nmkdir -p reports/kardashev-business\nnpm run --silent demo:kardashev-business:plan -- task energy foundation > reports/kardashev-business/order.json\nnpm run --silent demo:kardashev-business:plan -- example foundation > reports/kardashev-business/evidence.json\nnpm run --silent demo:kardashev-business:plan -- review reports/kardashev-business/evidence.json foundation\n```\n\nThe example produces `analysis.json` inside an evidence receipt. The checker verifies source/task/artifact hashes, exact byte counts, unique rows, all balances and the deficit set. It never grants production or settlement approval. A modified answer with a recalculated hash still fails arithmetic checking. A hash alone cannot authenticate the worker, reviewer, time of execution or economic provenance.\n\n## Five preserved simulation engines\n\nUse Python 3.12 from the repository root. Every command below is a bounded simulation; none invokes a live model or submits a transaction. Checkpoint and owner-control files belong to each engine. Prefer the isolated smoke command when comparing first runs without resuming old state.\n\n| Engine | Launch | Focus |\n| --- | --- | --- |\n| Foundation | `python -m demo.kardashev_ii_omega_grade_alpha_agi_business_3 --max-cycles 3 --no-resume` | Resource ledger, recursive tasks |\n| Business 3 | `python -m demo.kardashev_ii_omega_grade_alpha_agi_business_3_demo --cycles 3 --no-resume` | Integrity, worker health and validation windows |\n| Omega | `python -m demo.kardashev_ii_omega_grade_alpha_agi_business_3_demo_omega --cycles 3 --duration 3` | Mission planning and operator graph |\n| Supreme | `python -m demo.kardashev_ii_omega_grade_alpha_agi_business_3_demo_supreme --cycles 3 --no-resume --validator_commit_delay_seconds 1 --validator_reveal_delay_seconds 1` | Messaging, checkpointing and telemetry |\n| Ultra | `python -m demo.kardashev_ii_omega_grade_alpha_agi_business_3_demo_ultra launch --cycles 3 --no-sim` | Mission archives and owner controls |\n\n`npm run demo:kardashev-business:smoke` launches all five with temporary state and a per-process timeout. Foundation and Supreme now finish by default. Explicit zero-cycle options retain the documented indefinite mode; Omega also needs `--duration 0` to remove its time bound. Ultra's `--cycles 0` delegates to mission settings.\n\n## Work that can be specified today\n\nThe ten templates cover performance optimization, software features, SDKs, test suites, AI evaluations, dashboards, product comparisons, technical documents, energy analysis and numerical reproduction. Each exports the existing `ComputerWorkTask` schema: exact inputs, a worker profile, an origin allowlist, acceptance criteria and named deliverables. All included exercises use synthetic, non-personal inputs. Authorized public or licensed inputs can be used in separately reviewed tasks.\n\nThese are reviewable task specifications, not proof that every model can complete every task. Text deliverables can contain source code and build instructions; this adapter currently accepts text, Markdown, CSV and JSON, not arbitrary Office binaries or screenshots. Store additional screen/action evidence separately under the deployment's access and retention controls, and bind it into the independent review record.\n\n## Connect a qualified OpenClaw worker\n\nInstall repository dependencies using the pinned versions and lockfile before using the existing TypeScript adapter:\n\n```sh\nnvm install\nnvm use\nnpm ci\nnpm run demo:kardashev-business:worker -- inspect energy foundation\n```\n\nFor a downloaded or edited task file, use `npm run demo:kardashev-business:worker -- inspect-file /absolute/path/task.json`, then dispatch an independently admitted task with `npm run --silent demo:kardashev-business:worker -- run-file /absolute/path/task.json JOB_ID`. Editing any task field changes its admission digest. The included energy checker accepts only the exact bundled energy task; changed tasks need their own acceptance checks.\n\nThis inspection normalizes the task with `apps/orchestrator/computerWork.ts` and prints the exact digest. The workbench and adapter digest are cross-checked in tests for every template and family.\n\n1. Provision an isolated worker environment and an OpenClaw agent named `kardashev-business`. Enable the documented Responses endpoint only after its authentication and network boundaries are configured. Use separate trust boundaries and credentials for unrelated operators. Do not put a signer or wallet seed in the worker.\n2. Qualify required tools on that installation. OpenClaw browser automation, the OpenAI API computer-use loop and ChatGPT Work desktop Computer Use are distinct execution paths. A successful HTTP health check does not establish that a desktop session, app permission or browser action works. No direct ChatGPT Work bridge is implemented by this workbench.\n3. Copy `worker-profiles.example.json` outside the public repository. Replace the deployment identifier with the selected chain/contract identity. Configure an absolute persistent journal directory and a bearer token in `COMPUTER_WORK_KARDASHEV_TOKEN`; the profile contains only the environment-variable name. Do not publish tokens or local profile files.\n4. Configure runtime-enforced app/site/action allowlists, filesystem isolation, tool-step and wall-clock limits, provider budget limits, cancellation and consequential-action approvals. The task's origin list and instructions are not an OS sandbox. Adapter timeouts and output-token limits are not total-spend caps.\n5. Admit the **actual job ID and exact normalized task digest** in the profile's `approvedJobs`. The shipped profile has an empty list and therefore cannot dispatch. Set `COMPUTER_WORK_PROFILES_FILE` to your private profile and `COMPUTER_WORK_STATE_DIR` to the persistent journal directory.\n6. Run `npm run --silent demo:kardashev-business:worker -- run energy foundation JOB_ID` and save stdout as a private receipt. This is a real provider dispatch only when your admitted live profile is configured. It is not authorized by selecting a workflow on the website.\n7. Check the energy receipt with the CLI or browser, then obtain independent substantive review. For other templates, implement and run task-specific checks and independent review; the energy checker deliberately rejects their receipts.\n\nThe shared adapter binds dispatch to job ID plus task hash, uses an exclusive persistent journal, enforces response/time limits, and returns `productionApproved: false` and `settlementApproved: false`. On timeout or interrupted execution, preserve the journal and reconcile with the worker. Do not delete replay records or blindly retry an outcome that is unknown. Stopping the client does not prove the remote worker stopped: use the worker's cancellation control and verify its state.\n\n## Review and settlement boundary\n\n```mermaid\nflowchart TD\n A[Authorized work order] --> B[Isolated specialist execution]\n O[Owner controls] --> B\n B --> C[Artifacts and execution evidence]\n C --> D{Automated checks pass?}\n D -->|No| R[Revise within authorized limits]\n R --> B\n D -->|Yes| E[Independent substantive review]\n E --> F{Accepted and authorized?}\n F -->|No| H[Hold or dispute]\n F -->|Yes| S[Separate signer verifies chain and job state]\n S --> T[Settlement receipt and finality]\n```\n\nThe Python engines' synthetic token balances, simulated validators and toy planetary metrics remain simulation units. They are not live USDC flows. For real settlement, verify the actual deployed manager's chain, contract, token address/decimals, job state, acceptance rules and finality. No settlement adapter is added or silently enabled by this change.\n\n## Vision and honest capacity accounting\n\nThe website connects verifiable digital production to potential reinvestment in research, tools, compute and energy infrastructure. Kardashev Type II is a long-term civilizational ambition, not an attained capability, a deployment certification or a prediction.\n\nThe **$40 trillion/year** screen-work market is a user-supplied scenario assumption. **$40 billion/year** is the first major volume milestone and equals **0.1%** of that assumption. Neither figure is validated revenue. The explorer uses:\n\n- Worker capacity = agents × jobs/agent/day × useful-output fraction.\n- Review capacity = reviewer hours/day × 60 ÷ minutes/job.\n- Reviewable throughput = minimum of worker capacity, review capacity and the chosen market ceiling divided by assumed job value.\n- Illustrative annual gross volume = throughput × 365 × average USDC value, assuming $1/USDC.\n\nCosts, demand validation, retries, disputes, variable utilization and taxes are excluded. Physical deployment would require separate engineering, permissions, resources and real-world evidence.\n\n## Qualification and verification\n\n```sh\nnpm run demo:kardashev-business:test\nnpm run demo:kardashev-business:smoke\nnpm run demo:kardashev-business:qa\n```\n\nThe last command requires the repository Playwright/axe dependencies and Chromium (`npx playwright install chromium`). It checks desktop/mobile layouts, keyboard access, task downloads, imported evidence, deliberate corruption and accessibility. The Pages pipeline publishes the same local assets at `experiments/kardashev-business/`.\n\nProduction operation additionally needs representative live-provider and desktop/browser commissioning, prompt-injection and permissions tests, independently reviewed paid pilots, durable recovery, monitoring, cancellation and a verified settlement integration. Passing local tests cannot establish those outcomes. No live provider, real Mac, unrelated reviewer or paid settlement is claimed by this suite.\n\n## Current provider references\n\nDocumentation checked 2026-10-06:\n\n- [ChatGPT Work Computer Use](https://learn.chatgpt.com/docs/computer-use): supported desktop app interaction and app permissions.\n- [OpenAI API computer use](https://developers.openai.com/api/docs/guides/tools-computer-use): runtime execution, screenshots and bounded tool loops.\n- [OpenClaw browser tools](https://docs.openclaw.ai/tools/browser), [OpenAI provider](https://docs.openclaw.ai/providers/openai), and [security](https://docs.openclaw.ai/gateway/security): verify installed capabilities, authentication and gateway trust boundaries.\n","format":"text","sha256":"01bd9e214d8fd97e798eef104449399ecfde12cbc7c53e8caa3ac8c34c631c68","bytes":11573,"download":"/AGIJobsv0/examples/01bd9e214d8fd97e-README.md","source":"https://github.com/MontrealAI/AGIJobsv0/blob/5b4cebb309a83a7a6749d8911d8bf96a1921e042/demo/kardashev_ii_omega_grade_alpha_agi_business_3/workbench/README.md"}]}
FROM READING TO A REPRODUCIBLE RUN
Try the selected path.
Python, no demo package install
Run from the repository root with Python 3.12. This selected entry point uses the Python standard library. Keep output in a separate directory so that you can compare runs.
A bounded synthetic orchestration run. Use the workbench for task templates and exact artifact checks; live-provider execution and settlement require separate qualification.
The source inspector above reads bundled repository material. Local commands run separately on your computer. Recorded examples may contain historical timestamps, placeholders and simulated metrics.
MAKE IT YOUR OWN
One useful experiment.
Generate an energy-ledger artifact, test a deliberately wrong answer, and vary reviewer capacity before increasing the agent count.
THE SYSTEM, MADE VISIBLE
Architecture & relationships
This supporting entry has no standalone architecture diagram. Follow its related experience below to inspect the canonical system.
WHEN SOMETHING DOESN’T MATCH
Troubleshooting
Import or dependency error
Confirm the active virtual environment and the selected demo’s requirements. Run python -m pip check; do not install unrelated demo requirements over a working environment.
Unexpected result or missing file
Check the selected entry point, configuration and output argument. Keep the seed and implementation fixed before comparing outcomes.
TRACE THE CHECKS
Verification & next steps
2 tracked test source files are available in this directory. Inspect the tests and their environment before choosing a suite; file counts do not establish test results.