Oryn — local AI evaluation worksheet

Current status: developer alpha, Apple Silicon Macs. Public browser and SDK releases are not available yet.

1. Pick one task
What does the user want to accomplish? What inputs and outputs are required?

2. Set a quality bar
Choose representative examples. Define what a good result means before comparing models. Include difficult and failure cases.

3. Check device fit
Record Mac model, memory, model version, runtime and task settings. Measure first download, warm/cold response times and resource use. Test cancellation and insufficient resources.

4. Keep existing customers
Define the normal website experience in Chrome, Safari and other browsers. Specify your own fallback when Oryn is unavailable or permission is declined. Disclose any cloud processing.

5. Check the economics
Compare local adoption, cloud inference saved, hosting, implementation, support and any fallback costs. A local model avoids cloud inference fees for that work, not every business expense.

6. Explain privacy
Describe what runs locally and what your website receives, stores or uploads. Do not equate local inference with a completely offline or private website.

7. Make a decision
What did the results establish? Which users and devices can you support? What must improve before rollout?

This worksheet runs no model and collects no information. Fill it in using your own tools.