This is a test of Dark Factory, an AI-operated dark factory. Our flagship ETSS is preliminary: an experimental way to explore technical options with AI.
A dark factory automates the work normally performed throughout a service workflow. Here, AI carries out research, drafting and customer communication within defined limits. You are communicating with AI, not a person leading your engagement. Human ownership and operating limits remain; fully unattended operation is still being tested.
ETSS outputs are interim and unauthenticated. They are not certified engineering advice or approved designs and require appropriate independent professional review before engineering use.
SYNTHETIC METHOD DEMONSTRATION
A small cost advantage is not a selection.
Two fictional process-water filtration packages show how we trace assumptions, reproduce calculations, and test whether a conclusion survives a change in inputs. All numbers are invented. This example does not select real equipment or establish production readiness.
The result
At the assumed 4,000 operating hours per year and CAD 0.12/kWh electricity price, package B has lower included five-year cost by CAD 2,100. The source assumptions and full calculations are in the downloadable package.
| Illustrative case | A: included cost | B: included cost |
|---|---|---|
| Base case | CAD 88,800 | CAD 86,700 |
| 2,000 operating hours/year | CAD 74,400 | CAD 77,100 |
| CAD 0.08/kWh electricity | CAD 79,200 | CAD 80,300 |
These undiscounted five-year costs omit material real-world factors. Technical eligibility is assumed only for the arithmetic; actual performance, safety, operability, and full lifecycle costs remain unknown. There is no real technology recommendation.
Review and reproduce
- Start with the one-page summary (PDF)
- Read the two-page example (PDF)
- Download the inputs, source assumptions, Python reproduction, and checker findings (ZIP)
The reproduction uses Python 3 with no additional packages, account, API key, or paid service. The evidence package records the independent AI checker review. Neither AI agent supplies professional engineering approval.
Help us improve the next example
Could you follow the calculation? Which missing input would matter most to your decision? What narrow comparison should we demonstrate next?
Send optional feedback. Keep it generic and non-confidential. Downloading or sending feedback does not subscribe you to marketing emails.
For real project use, professional review, validation, authentication, and issue control belong with an appropriate Licensed Engineering Partner where required.