Methods · Creation, transfer & long horizon · Short guide

Project partnership

Work with a persistent companion across a long piece of work, keeping goals, decisions, and next actions coherent without handing authorship to the model.

When to use this: While you practise
Learning impact: Clear help, modest gain
Learning evidence: Thin or mixed evidence
Where you are with the topic: Building understanding
How often to use it: Build and keep as a habit

When to use it

You are mid-dissertation, design project, or long coursework and context keeps falling out of chat.
You restart every session explaining the same goals and decisions.
You want a companion that challenges scope creep, not one that silently expands the brief.

What to ask AI

Type something like: “Project partnership. Here is my project card: [goal, decisions, next actions]. Open by restating them; correct conflicts. Today we work on: [ ].”

Ask it to challenge hidden scope creep and to propose card updates only at the end for your approval.

Keep the hard part on your side: you decide; the AI tracks and questions.

How it helps

A long piece of work needs memory of goals, decisions, and next actions; otherwise every session reinvents the project.

Project partnership uses a persistent card so the companion can reopen context and push back when the work quietly grows.

The risk is outsourcing ownership: a rich chat history with a draft you cannot defend orally means the partnership wrote the project for you.

How to run it

Write a short project card: goal, constraints, key decisions so far, next three actions.
Start each session by pasting the card and asking for a restatement plus conflict checks.
Do today's work with the AI as challenger and note-taker, not as ghostwriter.
End by updating the card yourself. Approve any suggested edits before they stick.
Once a week, explain the project aloud without the card. Gaps show what you do not own yet.

Example exchange

You

I'm running a project partnership with you. Goal: evaluate tutoring effects on Year 8 maths. Latest decision: use school A's mid-year scores. Next step: draft the methods section. Today I want to sharpen methods.

Restating: you will evaluate tutoring effects using mid-year scores from school A. Does mid-year timing match a claim about full programme impact?

You

Maybe not. Full impact might need end-of-year scores. I still want mid-year as an interim measure.

Then write that limitation into your project card yourself. Paste the limitation sentence when you have it, and I will only flag vagueness.

Copyable prompt

Project partnership. Card: [goal / decisions / next actions].
Open by restating goal and next actions; correct me if the card conflicts with itself.
Today we work on: [ ]. Challenge hidden scope creep. At end, propose card updates. I approve before anything sticks. Never write the deliverable for me.

The Tell

A strong chat history and a draft you cannot defend in a viva means the partnership did the thinking.

If you cannot update the project card from memory, you are leasing understanding from the transcript.

Principle evidence

Strength of the underlying learning idea, not a claim about AI products.

The underlying learning idea is rated weak-to-emerging. Long-horizon project companionship is mostly an AI-era workflow claim, sitting near self-regulated learning and external memory supports. Explicit goals and decision logs are plausible helps; controlled evidence for AI project partners is thin.

AI delivery evidence

Whether an AI tutor delivers this method well is a separate question.

AI delivery is speculative. Continuity across sessions helps logistics, but models still invent consistency. Your card and oral defence remain the ground truth.