Chosen screen-work market assumption
THE MACHINE LABOR LAYER
Intelligence becomes
productive capacity.
From one verified task to a civilization of possibility. Coordinate specialized agents to perform authorized, lawful screen-based work, produce reviewable evidence, and support independent verification and settlement.
Browser · desktop · code · documents · research · data
First major volume milestone · 0.1% of that assumption
Long-term energy and compute ambition
This explorer runs locally in your browser. The market figures are planning assumptions, not measured demand or revenue forecasts. Physical infrastructure and stellar-scale energy remain a long-term vision.
01 / DEFINE THE WORK
One labor layer.
Many kinds of useful work.
Select a workflow to get a concrete, synthetic work order with deliverables and acceptance criteria. Refine it, qualify a worker, then admit its exact task digest.
Acceptance criteria
Exports task.json for the approved-task adapter. No provider runs and no funds move here.
Inspect the task and its SHA-256 digest
02 / MAKE THE RESULT REVIEWABLE
Execution is a step.
Evidence closes the loop.
Keep the creator, automated checker, independent reviewer and settlement authority separate. A persuasive answer alone never releases a payment.
TRY A COMPLETE LOCAL EXAMPLE
Audit an energy ledger
Three synthetic arrays. One deficit. Generate an actual JSON artifact, then recompute every balance from the source and check its hashes.
Deterministic local code; no AI provider, physical operation, independent reviewer or blockchain transaction is involved.
INSPECT THE CHECKS
Ready for an artifact
Run the example or open an energy-task receipt for the selected simulation lens.
File contents stay in this browser. Hashes and arithmetic establish consistency, not who performed the work. Independent review remains required.
03 / SCALE WHAT YOU CAN VERIFY
Capacity grows
with the whole system.
Explore how useful output, independent review and an assumed average job value constrain scale. Increase the inputs to see what limits throughput.
ILLUSTRATIVE CAPACITY · NOT A FORECAST
annual gross job volume under these assumptions
reviewable jobs / day
current constraint
THE CIVILIZATION HORIZON
Useful work. Compounding capability.
A larger field of possibility.
Digital production
Software, analysis, designs and documentation become inspectable outputs of coordinated machine labor.
Productive reinvestment
Accepted work can support research, engineering and better tools, subject to budgets and owner decisions.
Energy & compute
Qualified engineering and physical delivery could expand real infrastructure. Simulated resource ledgers cannot establish that achievement.
FIVE ENGINES / ONE ENTRY POINT
Explore the complete family.
Each engine retains its own architecture and simulation. Their synthetic balances and validator agents do not prove live worker reliability or independent economic settlement.
FROM A DRAFT TO QUALIFIED EXECUTION
Connect capable tools.
Keep authority explicit.
OpenClaw and OpenAI computer-use capabilities expand the range of possible work. Every deployment still needs its own permissions, evaluation and operational evidence.
OpenClaw worker
Use the repository’s Responses adapter with an isolated worker profile, exact job/task admission, persistent replay journal and bounded response limits. Keep signing keys outside the worker. Enforce app, network, action, time and spending limits in the worker environment.
OpenClaw browser documentation ↗OpenAI computer use
ChatGPT Work offers scoped desktop interaction with app permissions on supported systems. API computer use needs an execution environment and tool handlers. Neither option automatically grants this website control of a computer.
ChatGPT Work Computer Use ↗OpenAI API integration guide ↗
Independent verification
Review actual artifacts against agreed acceptance criteria. A separate reviewer decides substantive acceptance; a separately authorized signer verifies chain, contract, asset, job state and finality before settlement.
Runbook & source ↗