Skip to content
Retri

ChatGPT workflow

How to use spaced repetition with ChatGPT

A useful ChatGPT conversation can leave you with a clear explanation and very little practice recalling it. Add a short review step: pick what matters, turn it into questions, and come back to those questions later.

Begin with something you want to be able to do

“Learn SQL” is a large project. “Explain when I’d use a LEFT JOIN instead of an INNER JOIN” is a question you can work on today. A clear goal helps you decide what belongs in your deck and what can stay in your notes.

Ask ChatGPT for an explanation, then work through an example yourself. If you’re still confused, keep discussing the concept before making cards. A polished answer on the back of a card won’t resolve a gap in understanding on its own.

Use this prompt to narrow a session:

I want to understand LEFT JOIN versus INNER JOIN. Explain the difference using two small tables, then ask me to predict the output of a new example.

Choose a few ideas worth keeping

At the end of the conversation, look for ideas you expect to need again: a distinction you keep mixing up, a term you need to explain, or a decision you want to make without reopening the chat.

Skip details you can easily look up and won’t use often. Your deck should reflect your learning goal, not the length of the assistant’s response. Our guide to information overload offers a filter for deciding what to save.

Suggest up to five flashcards from this conversation that would help me choose and explain a SQL join. Show the drafts first. Keep one clear question per card and leave out trivia.

This asks for drafts. Read them before asking Retri to save them.

Check the answers before you practice them

ChatGPT can produce mistakes or leave out conditions that matter. Compare technical claims with the original documentation, course material, or another authoritative source. For the join example, verify the behavior in your database’s documentation and run the query against sample data.

Revise answers that depend on words like “always” or “never.” Put the relevant version, context, or exception in the question when it changes the answer. If you can’t explain why an answer is correct, investigate before adding it.

For more examples, see how to write better flashcards from AI conversations.

Save the cards with the Retri ChatGPT plugin

Install Retri in ChatGPT, then connect your Retri account. Once connected, ask ChatGPT to create a deck and save the drafts you approved. The setup guide covers connection steps.

Create a Retri deck called SQL Foundations. Use one question per card, concise answers, and a small example when it clarifies the idea. Save the flashcards I just approved to that deck.

Retri stores your cards and schedules reviews. It doesn’t automatically save every conversation; you choose what to ask it to keep. You can edit, suspend, or delete cards as your needs change.

Attempt the answer before revealing it

Ask your connected assistant to start a review:

Review my due cards in the SQL Foundations deck.

Read the question and answer from memory. Then reveal the card and compare. Use Again if you couldn’t recall it, and choose Hard, Good, or Easy to reflect a successful answer’s difficulty. Retri uses your ratings to schedule future reviews.

The value of the recall attempt has research behind it. In Roediger and Karpicke’s 2006 experiments with prose passages, prior recall tests led to better delayed retention than repeated study, even though repeated study made learners more confident. Read the study.

Finish by using what you recalled

After reviewing the join cards, write a new query. Explain which rows you expect to keep and why. If your explanation fails on a new example, ask for help and revise the relevant card.

This gives the next conversation a purpose: address a specific gap, rather than collect another broad explanation. Over time, your deck becomes a record of the knowledge you’ve chosen to practice.

A workable first session can be simple: learn one concept, approve a few accurate cards, save them, and return when they’re due. Read what SRS means if you want the method behind the schedule.

AI-assisted guide from Retri. Claims were checked against the linked sources and current product behavior. Examples are illustrative. Learning research cited here evaluates study methods, not Retri itself.