Choose one of three approaches. The steps for that option appear below.
Choose an option above. On a phone, swipe the tabs left or right to see all options →
ChatGPT proposes episodes and builds the journal table. You check every episode and test the journal.
Store sets locally
Draft episodes
Check episodes
Build journal table
Keep and test
Use: You
Store the 5 marked maths practice sets, the authorised corrections and your reflections locally.
Use: ChatGPT
Attach the practice-set extracts and ask ChatGPT for an episode row per error, with candidate links to earlier episodes of the same type.
What to paste into the tool
Task: Draft error episodes with correction text and uncertain repeat links. Here are my five marked practice sets, the authorized corrections and my reflections. For every error, create an episode row: the set and question, what I did wrong in my own working, the authorized correction, my reflection, and candidate links to earlier episodes of the same error type. Mark uncertain links as candidates, not facts.
Use: You
Compare every episode with the retained working and authorised corrections; fix omissions and accept or reject repeat links.
Check repeat links without inventing a diagnosis; keep unresolved links marked.
Use: ChatGPT
Ask ChatGPT to structure the checked episodes into a journal table: one row per episode with date, set, topic, error, correction, cue and links.
What to paste into the tool
Task: Candidate error journal table with checking cues and recurrence links. Structure the confirmed episodes into a journal table I can keep and search: one row per episode with date, set, topic, my original error, the checked correction, a short check cue I can run on my next attempt, and links to earlier episodes of the same type. Output it as a plain table I can paste into a spreadsheet or notes app.
Use: You
Keep the journal in a local spreadsheet or notes file; test search, inspect recurrence history, try each cue on fresh practice. Completion is not mastery.
Check search finds a known episode and corrections match the originals; the journal is untested until each cue works.
Gemini Notebook proposes episodes from your sources and generates the table. You write the ledger, export and test.
Store sets locally
Draft episodes
Write the ledger
Generate table
Export and test
Use: You
Store the 5 marked maths practice sets, the authorised corrections and your reflections locally.
Use: Gemini Notebook
Attach the extracts to a notebook and ask for every error episode with citations to the original working, correction and reflection, plus candidate repeat links.
What to paste into the tool
Task: Episode worksheet requesting citations to original working and corrections. Using only the sources in this notebook, extract every error episode from the five practice sets. For each episode cite the original attempt, the authorized correction and my reflection, and propose recurrence links to earlier episodes with citations to those earlier attempts. Mark uncertain links as candidates.
Use: You
Compare every episode with the working and corrections; write a recurrence ledger and choose a check cue per episode.
Check repeat links without inventing a diagnosis; every episode gets a cue.
Use: Gemini Notebook
Add the ledger as a source and ask for a Studio Data Table: one row per episode, cues and confirmed links kept verbatim.
What to paste into the tool
Task: Candidate error journal table with checking cues and recurrence links. Use the screened ledger as a selected source; preserve learner-written checking cues and confirmed links verbatim; use one row per episode with date, attempt code, original working, correction, cue and recurrence reference; retain uncertainty labels
Use: You
Export the table and citation tab to Sheets; keep locally, test search, inspect recurrence history, try each cue on fresh practice. Completion is not mastery.
Check export permissions, rows and the citation tab; the journal is untested until the search finds a known episode.
You transcribe the marked work yourself. Gemini organises the episodes and proposes cues and links. You test the journal.
Type the episodes
Build journal table
Keep and test
Use: You
Type episode IDs, dates, original working, authorised corrections and reflections from the 5 marked sets into your own worksheet, keeping unclear entries.
Check each typed row matches the marked set and the correction it came from.
Use: Gemini
Attach the checked episode list and ask Gemini for a table: one row per episode with date, set, topic, error, correction, cue and links.
What to paste into the tool
Task: Candidate error journal table with checking cues and recurrence links. Here is my confirmed list of error episodes from five practice sets. Structure it as a journal table with one row per episode: date, set, topic, my error, the checked correction, a check cue I can run on my next attempt, and proposed recurrence links between episodes of the same error type. Mark the links as candidates for me to confirm.
Use: You
Keep the journal in a local spreadsheet or notes file; test search, inspect recurrence history, try each cue on fresh practice. Completion is not mastery.
Check the search finds a known episode and corrections match; the journal is untested until each cue works.
Best when: You want ChatGPT to propose episodes and the journal layout
What it does: ChatGPT drafts episodes and the table; you check and test
Strong at: Documents and spreadsheets listed as conversation uploads
Limit: Episode completeness and cue usefulness not established; test first
Free plan: Yes, limited (uploads, images, voice, data analysis)
Best when: You want source-grounded episodes and a Sheets export
What it does: Gemini Notebook drafts and tabulates; you write the ledger and check
Strong at: Chat answers cite your sources; export keeps a citation tab
Limit: Episode completeness and cue usefulness not established; test first
Free plan: Yes, limited (notebook and source caps)
Best when: You want to transcribe the work yourself, Gemini organises
What it does: You transcribe; Gemini organises and proposes cues and links
Strong at: Answers based on your uploaded content
Limit: Episode completeness and cue usefulness not established; test first
Free plan: Yes, limited
Before you paste into any AI tool, remove any confidential information. Put it back at the end.
Deciding between ChatGPT and Gemini for this? See the comparison.
Keep the journal in a local spreadsheet or notes file, test search with a known episode ID, verify original attempts and checked corrections, inspect recurrence history and try each cue on fresh practice while distinguishing journal completion from mastery.
Check the result against your original inputs. Approve it before sharing or acting.