Open research

What changed,
and how did we measure it?

REPROF’s research program is designed to make its own claims harder to fake: pre-registration, comparable baseline/endline conditions, external scoring and publication of null or negative results.

Founding hypotheses

REPROF is testing whether AI can reduce prerequisite knowledge and time required to begin meaningful work in selected domains, and whether a person can become better at supervising AI rather than merely faster at accepting it.

Primary constructs

  • Verification discipline
  • Error detection
  • Confidence calibration
  • Recovery from misleading hypotheses
  • Scope and accountability
  • Communication
  • Orchestration judgment

Comparison design

Day-0 and Day-30 assessments should use parallel forms under equivalent access conditions. Waitlisted non-enrollers may be invited to take the same forms at the same interval as a non-randomised comparison group. This does not establish causality and must not be described as an RCT.

Failure criteria

The research only matters if failure is publishable. REPROF intends to report completion, dropout, zero or negative deltas, cases in which AI made performance worse and limitations in grader agreement.

Cohort Zero

The first independently organized NEW PROFESSION #01 cohort is an early research input. It is not proof of the global model, and global development does not depend on any single local cohort or commercial result.

Pre-registration status: draft instrument architecture exists; the final pre-registration, task battery and external reviewer plan must be locked before the first scored use.