KNOW WHAT WAS CHECKED
Evidence you can inspect.
Documentation, user reports and actual tests answer different questions. We keep those labels separate.
| Label | What it establishes |
|---|---|
| Documentation reviewed | A dated source describes the capability. It has not been run by us. |
| User reported | A reader describes their own result. We review the submission; it remains their report. |
| Maintainer confirmed | A correction is supported by a public identity link and reviewed evidence. This is not a ranking boost. |
| Tested by us | A published record states the version, environment, procedure, input scope, date, result and limitations. |
A useful first trial
- Choose a small input you have permission to use and define the expected output.
- Record the version, OS, architecture, dependencies and model settings.
- Run in an isolated environment. Record any network services and costs.
- Check the result manually, including at least one difficult or failure case.
- Report failures and limitations alongside success. Keep secrets and private inputs out of shared evidence.
We have not published hands-on tests of the listed projects. Suggested trials in guides are procedures you can use, not completed benchmarks.
Share a setup experience or correction →Reviewed public reports
No reviewed public reports have been published yet. Submissions never appear automatically.