The write-to-learn note is consolidation, done the same day I finish a topic. I get an LLM to generate a sizeable prompt that spans the whole topic, then answer it in my own words — roughly 500 words, no copy-paste. Explaining it forces the gaps to the surface, and I capture those at the bottom so the active-recall note
It is not tracked by the dashboard — it has no stage. It’s the raw material the spaced reps are built from.
Template
Copy this into a new note (e.g. DD-MM-YYYY-Topic.md):
#write-to-learn #cloud #aws #topic-tag
*"[AI-generated prompt: one or two sentences that span the whole topic and name
the things you should be able to explain. Ask the model to be specific and to
include the comparisons/edge cases it expects you to know.]"*
**[Sub-topic heading]:**
[~500 words answering the prompt. Define technical language as
you use it. Where you compare two things, lay out the trade-off explicitly.]
---
##### Knowledge gaps
- [one line per thing you couldn't explain cleanly — these become Q&A and flashcards]How to use it
- Finish the material (a course section, a doc, an incident write-up).
- Ask an LLM for a single prompt covering the topic — “what should I be able to explain about X? Phrase it as one dense question.” Paste it in italics at the top.
- Answer it in prose from memory first, then open the source to fix what you got wrong. The act of writing while slightly unsure is where the learning is.
- Every time you stall, write the gap in the Knowledge gaps list.
- Promote the topic to an active-recall note and (optionally) run the flashcard generator over this write-up.