Once you’ve noted a question in LectureQ, it’s time to find the answer. These days, most of those answers come out of a conversation with AI. The catch: two people can ask the same question and walk away with very different things. This guide follows LectureQ’s flow (question → answer → what I learnt) and sets out how to make that difference work in your favour.
The one skill this guide sets out to build:
Turning your questions into ‘verified knowledge’ with the help of AI.
— Not the ability to receive answers, but the ability to doubt the answers you get and make them your own.
One question, one conversation. Cram several questions into one chat and the answers blur together — and you won’t be able to sort them out later.
A question scribbled down in a hurry during a lecture is usually fuzzy, and fuzzy questions get fuzzy answers. Don’t ask for the answer straight away — rewrite the question first.
Below is a question I jotted down in a hurry during a lecture. It's still fuzzy.
"{the question exactly as written in LectureQ}"
Rewrite it as 3 questions that can actually be answered.
For each one, add a line on what I'd be able to do once I know the answer.
Don't answer yet — just show me the rewritten questions.
▸ When you get a version you like, update the question in LectureQ to match. It’s an investment in your future self.
The AI has no idea how much you already know. Just adding where the question came from / what you know right now / what kind of answer you want makes a striking difference to the quality of the answer. All three are already written down in the app.
[My question]
{the refined question}
[Context]
- Where it came from: {lecture name} / {the time-block notes from that point}
- What I know right now: {"Where I Am Now" from Learning Status}
- Where I want to get to: {"Where I Want to Be When Done" from Learning Status}
[The answer I want]
1) The core answer in 3–5 lines
2) One paragraph on why
3) 1–2 things I'm likely to have misunderstood
4) The smallest example or experiment I can try myself to check it
Mark clearly what's certain and what's a guess. If you don't know, say so.
▸ Where to find the context: View → Learning Status in the toolbar, and the notes of the time block the question sits in.
A plausible answer is not a verified answer. Push back at least once.
Which part of your answer is most likely to be wrong? - The assumptions that must be true for this answer to hold - The opposing view, or exceptions where this answer doesn't work - What should I look at to check this against primary sources? (names of official docs, specs or original texts)
Information like versions, figures, API names and years — the more specific it is, the harder it is to notice when it’s wrong — should always be checked against the original source.
A good answer always uncovers something new you don’t know. That’s your next question — what the app calls a follow-up question.
Based on our conversation so far, give me 3 questions about things I don't know yet but should. No explanations, just the questions. Short and specific.
▸ Paste those questions straight into the ‘New follow-up question’ field in the question details. They’re kept as original question → follow-up, and the Graph shows how your understanding has branched out.
Copying and pasting the AI’s answer leaves a record, but no knowledge. Write the Answer field in your own words, then have those sentences graded.
Let me put what I've understood into my own words.
"{the summary I wrote myself}"
Point out anything wrong, missing or overstated.
Instead of a score, show me a corrected version.
▸ Only mark something Resolved + What I Learnt (★) once it has made it this far. The ‘What I Learnt’ list in View → Summary is only worth something when it holds nothing but what you’ve filtered this way.
Here are the five steps above, put in terms of skills. Start practising the row you feel weakest at.
| Skill | I can… | I know about / Related keywords |
|---|---|---|
| Refining questions | Rewrite a fuzzy note as three answerable questions | What makes a good question, answerable vs. unanswerable questions, problem framing |
| Giving context | Ask with background, my current level and the answer format I want | Structured prompts, specifying role, context and output format, context design |
| Checking answers | Sort fact from fiction by pressing on an answer’s assumptions, counterexamples and primary sources | Hallucination, evidence and sources, falsification, confirmation bias |
| Exposing what I don’t know | Pull new questions out of an answer and keep them as follow-ups | Metacognition, knowledge gaps, question trees |
| Locking in knowledge | Rewrite it in my own words, get it graded and confirm it as ‘What I Learnt’ | The Feynman technique, retrieval practice, spaced repetition |
Save a conversation with an AI as Markdown and the app turns it into a question tree (toolbar lecture menu → Import from Markdown…).
# This line becomes the lecture title ## Topic (the first question here becomes the original question; the ones after it become follow-ups) **Me** — The question I asked goes here **AI** — The answer I got goes here (multiple lines and code blocks are fine)
Once imported, each question has its answer filled in and is marked resolved. Then, in the Summary, keep only what you truly learnt with ★ and ‘Exclude’ the rest — and a whole conversation becomes your list of what you’ve learnt.
Don’t ask AI for answers — use it as a partner to test your questions. You ask the questions, you do the doubting, and you write the final sentence. LectureQ is where that process is kept.