Update — New direction: handwritten cards are now first-class, AI is the Copilot
It’s been a while since I first posted this thread. The original framing was “AI generates memory challenges,” but after several releases I want to make the new direction clear.
EngramQuest is no longer an AI generator. It’s an Obsidian-native spaced-repetition system where handwritten cards and AI assistance are both first-class, and AI has shifted from “generator” to “Copilot.”
Write your own cards — five formats, freely mixable in one note
Add a #flashcards/topic tag and the plugin picks them up automatically. You can write cards wherever it fits your workflow.
A simple example:
#flashcards/learning
Q: What is spaced repetition?
A: Reviewing right before you forget — the most efficient way to convert short-term memory into long-term retention. The interval grows automatically after each successful recall.
{{c1::Spaced repetition}} is one of the most effective long-term memory methods
Pythagorean theorem :: a² + b² = c²
The five supported formats are: Q:/A: multi-line for everyday use (answers can span multiple lines, include images, tables, code); — fenced Q&A for existing notes (wrap a card with — lines so blank lines inside are preserved); %%card%% long answer (safe mode for pasting full AI responses, explained below); {{c1::}} Cloze (Anki-compatible cloze syntax); and :: one-liner for quick concept pairs.
AI as Copilot — paste a full AI response as a single card with %%card%%
This was the biggest workflow pain point. You ask AI to explain something, AI returns a long response with separators, tables, code blocks, lists, multiple paragraphs, and you used to have to manually reformat it into a Q/A. Not anymore.
Example:
%%card%%
Q: How do you explain agentic testing?
A:
Agentic testing checks whether an AI system can reliably complete a task, not just whether a function returns the right value. Give the agent a realistic task, verify the final artifact, and check logs, tool calls, and failure recovery.
%%card%%
Everything between the two %%card%% lines is treated as a single card’s answer. Separators, blank lines, code blocks, and tables are all preserved. Paste a ChatGPT or Claude response in and you barely need to edit anything.
Fastest way to type the fence: on an empty line, press Ctrl+/ (or Cmd+/). Obsidian inserts %% %% and places the cursor in the middle. Type “card” and you’re done.
Edit cards during review — changes write back to your source note
If you spot a card that needs improvement mid-review, use the Edit, Highlight (==), or Blockquote (>) tools in the review session. The change is written directly back to your original note, so your note stays the single source of truth.
AI-generated cards take a different path. They live under engram-review/ai-cards/, so AI never touches your original markdown.
Achievements (1.8.0)
This version adds a 6-tier Daily Review progression: Focused Start (20 cards), Memory Surge (50), Sprint Breakthrough (80), Hundred Fever (100), State of Flow (150), and Memory Domain (200) in one day. Newly earned achievements now show on the completion screen the moment a session ends, with their icon and rarity glow. There are 10 milestone achievements in total, each with a 3D-rendered icon and three rarity tiers (UC, R, LEG).
Other improvements since launch
Image cards can now be embedded inside any card format using image.png or(path), and they render in the review session. Quest Map supports image occlusion, which masks regions of your vault images as questions. Full Dark Mode is now in: a custom deep-space floating-island scene replaces the light background, not just a darkened version of it. Multiple AI tools are supported — Claude Code, Gemini CLI, Cursor, and Codex — pick whichever you use, no lock-in.
If this helped you — or even if it didn’t —
I’d really appreciate:
A star on GitHub:
https://github.com/bahfahh/engram-quest
A rating on ObsidianStats:
https://www.obsidianstats.com/plugins/engram-quest
Feedback and bug reports welcome — reply here or open a GitHub issue.
Quest Map — Your notes become gamified island stages. Each node is a chapter; AI generates challenges (multiple-choice, ordering, matching) based on your actual content. You can’t advance until you’ve proven mastery. Boss battles gate the end of each chapter.
Memory Map — AI analyzes your note relationships and builds a visual knowledge graph in Obsidian’s native Canvas. Concepts are chunked, contrasted, and anchored — the way memory actually works.
Review Deck — Powered by the FSRS algorithm, it surfaces cards at the exact moment before you’d forget them. When you’re stuck, AI provides three levels of hints: clue → vault context → direct answer.
