You use one assistant to turn a week of readings into practice questions. Later, you want another assistant to critique your explanation of a difficult concept. The first session also contains suggested answers, a citation you have not found, and wording close to what your assignment asks you to submit.
Before moving anything, open the course policy.
Start with what the course permits
Instructor and institution rules decide whether an AI may help with research, study, outlining, revision, coding, or submitted work. The rules can differ by assignment. Write the applicable boundary into the briefing:
- AI allowed for private study only;
- AI allowed for feedback but not drafting;
- use must be disclosed or cited in a specified way;
- particular sources, tools, or data cannot leave the course system;
- no AI assistance permitted.
If the assignment forbids the use, stop. RelayLink approval does not override the policy, and moving work between two models does not change its origin.
Split allowed tasks without outsourcing authorship
Where the rules permit it, give each assistant a narrow role:
- Research support: suggest search terms or help organize sources you then locate and verify.
- Study: generate practice questions, compare your notes with a rubric, or explain a concept from permitted material.
- Revision: critique an argument you wrote, identify unsupported steps, or test whether your explanation answers the prompt.
Keep your attempted reasoning in the handoff. If the next assistant sees only a polished answer from the first, it cannot help diagnose what you understand. Ask for questions, critique, or error detection instead of a replacement submission.
Carry sources and uncertainty honestly
A student briefing can name:
- the learning objective and exact scope;
- the course’s AI rule for this task;
- assigned readings by title, author, edition, chapter, or stable URL;
- citations you personally verified;
- your current explanation or attempted solution;
- where you are uncertain;
- feedback already accepted or rejected, with why;
- the next study or revision question.
Never preserve an invented citation because it looks plausible. Check authors, title, publication details, page, quotation, and relevance against the original. Cite sources using the course’s required form. Do not imply that model-produced wording or reasoning is entirely human-authored.
RelayLink carries a text briefing, not a textbook, learning platform, notebook, code project, or browser tab. The destination assistant needs its own authorized access to any material it must inspect.
Use a self-handoff for your own study
Multiple assistants can connect to the same RelayLink account. Address the package to your own email, then draft, review, and confirm. The model never sends autonomously. Check that the package follows the assignment rule and does not include another student’s work, protected assessment material, personal data, or answers you intend to misrepresent.
The self-addressed package needs no contact request or second account and sends no email or magic-link page. It remains unread until one of your assistants pulls it; you can inspect it through the account portal. RelayLink does not synchronize sessions. It carries only the briefing you approved. The self-handoff guide explains the mechanics.
Treat the second assistant as practice, not proof
Two models can share the same misconception, derive the same wrong answer, or invent matching references. Verify against assigned sources, calculations, experiments, instructor guidance, and your own understanding.
Provenance labels help record whether wording was yours or model-drafted, but they do not decide what your institution considers authorship. The provenance guide explains RelayLink’s labels; your course policy still decides what you may submit and how you disclose assistance.
When this workflow is a poor fit
Do not use it for a closed-book assessment, prohibited assignment, restricted research data, or any task where the destination provider lacks approval. It is also a poor fit when the learning goal is to perform the reasoning unaided.
If a safe handoff would contain the answer but not your thinking, keep studying directly. The point is to preserve a legitimate learning state across your own assistants, not to create a cleaner route around the work you were assigned.