About / Research
Why AgencyThread exists
AI changed the meaning of “authorship” long before our institutional habits caught up.
When an AI system drafts language, develops alternatives, restructures an argument, or performs analysis, asking who physically typed the final words becomes less informative.
But the human role still matters.
AgencyThread was created to make that intellectual contribution visible.
Authorship isn’t who typed the words. It’s who shaped the work.
Current status
A working product and research program in private Alpha
AgencyThread is currently in Private Alpha. Human/expert validation for stronger or high-stakes claims has not yet been completed, and the Collaboration Profile remains experimental.
This site supports owner review and later comprehension and customer-discovery testing. It does not represent a public launch, final trademark clearance, certification product, or high-stakes instrument.
Evidence status
What we have done
- Built and scored a large synthetic case set with known intended answers, covering a wide range of human-agency and evidence-quality patterns, to calibrate the evaluation approach before applying it to real work.
- Deliberately tested the evaluation approach against difficult, ambiguous, and adversarial cases designed to find where it could be wrong, not only cases designed to make it look good.
- Checked for internal consistency and reliability before treating the current approach as ready for real use.
- Opened the Private Alpha itself: selected pilot participants testing the product on their own real work.
What we have not done
Not yet claimed
- Independent human or subject-matter-expert validation of AgencyThread’s results against outside judgment.
- Any statistical validation suitable for high-stakes, disciplinary, or credentialing decisions.
- Peer-reviewed publication of this work.
This is why AgencyThread remains in Private Alpha, and why the site avoids describing any result as validated, certified, or proven.