Explain this codebase like I start on Monday
Produces the orientation doc a new engineer actually needs: the three things that will confuse them, the load-bearing files, and the conventions nobody wrote down.
Builds a one-hour path through a concept you keep bouncing off: the minimum prerequisite, the idea itself, the two places your intuition will be wrong, and a test.
Teach me this in about an hour: {{IDEA}} WHERE I'M STARTING FROM: {{BACKGROUND}} WHY I NEED IT: {{PURPOSE}} Build the session in this shape: **1. The one prerequisite.** The single thing I must understand first, explained in a paragraph. If I already have it from my background, say so and skip. **2. The idea itself.** Explain it once concretely, using an example from my own domain based on what I told you. Then explain it once more abstractly. Concrete first, always. **3. Where my intuition will be wrong.** Two specific misconceptions someone with my background reliably has, why they're tempting, and what actually happens. **4. The boundary.** When this idea stops applying, and what replaces it. **5. Ten minutes of practice.** One exercise I can do right now with what's on my machine. Give the task, not the answer. **6. Check.** Three questions. If I can answer them, I have it. Include what a wrong answer to each reveals. Rules: - No analogies to trains, restaurants, or libraries unless the mechanism genuinely matches. - Do not flatter the difficulty ("this trips up a lot of people!"). Just teach. - If the idea is genuinely too large for an hour, say so and teach the most load-bearing third instead of skimming all of it.
A prompt is a skill with the serial numbers filed off. Install it once and your agent reaches for it on its own.
Paste this into ChatGPT, Claude Code, Codex, Cursor, or whatever you use:
Install the learn-one-hard-idea skill globally from https://promptsbuddy.com/prompts/learn-one-hard-idea/skill.md.claude/skills/learn-one-hard-idea/SKILL.md
---
name: learn-one-hard-idea
description: "Builds a one-hour path through a concept you keep bouncing off: the minimum prerequisite, the idea itself, the two places your intuition will be wrong, and a test. Use when the user asks for help with learning tasks like learning, teaching, study."
license: CC-BY-4.0
metadata:
source: https://promptsbuddy.com/prompts/learn-one-hard-idea
author: "Meghna"
version: "1.2"
---
# Learn one hard idea properly in an hour
Builds a one-hour path through a concept you keep bouncing off: the minimum prerequisite, the idea itself, the two places your intuition will be wrong, and a test.
## Inputs to collect first
Ask the user for anything below that they have not already given you. Do not
invent values for these.
- `IDEA` — The concept you keep failing to hold onto. (e.g. Why database isolation levels matter in practice.)
- `BACKGROUND` — What you already know, honestly. Being specific here changes everything. (e.g. I write application code against Postgres daily. I know what a transaction is. I have never tuned one.)
- `PURPOSE` — Why you need it — this decides the depth. (e.g. So I can review a colleague's change to our payment writes without nodding along.)
## Instructions
Teach me this in about an hour: {{IDEA}}
WHERE I'M STARTING FROM: {{BACKGROUND}}
WHY I NEED IT: {{PURPOSE}}
Build the session in this shape:
**1. The one prerequisite.** The single thing I must understand first, explained
in a paragraph. If I already have it from my background, say so and skip.
**2. The idea itself.** Explain it once concretely, using an example from my own
domain based on what I told you. Then explain it once more abstractly. Concrete
first, always.
**3. Where my intuition will be wrong.** Two specific misconceptions someone with
my background reliably has, why they're tempting, and what actually happens.
**4. The boundary.** When this idea stops applying, and what replaces it.
**5. Ten minutes of practice.** One exercise I can do right now with what's on my
machine. Give the task, not the answer.
**6. Check.** Three questions. If I can answer them, I have it. Include what a
wrong answer to each reveals.
Rules:
- No analogies to trains, restaurants, or libraries unless the mechanism genuinely matches.
- Do not flatter the difficulty ("this trips up a lot of people!"). Just teach.
- If the idea is genuinely too large for an hour, say so and teach the most
load-bearing third instead of skimming all of it.
## Notes from people who use this
- Be honest in BACKGROUND. Overstating it gets you an explanation you'll nod through and forget.
- Do the exercise before reading section 6. Reading the check questions first turns them into hints.
- Save section 3 somewhere. Misconception lists are the part worth rereading in six months.
---
Learn one hard idea properly in an hour · by Meghna · v1.2
From PromptsBuddy — https://promptsbuddy.com/prompts/learn-one-hard-idea
Licensed CC BY 4.0.
Every prompt is also available at /prompts/learn-one-hard-idea/skill.md — see all install options.
Every good explanation runs concrete → abstract, and almost every generated one runs the other way, because definitions are easier to produce than examples. The prompt fixes the order explicitly, and asks for the concrete example to come from your domain, which is what makes it stick.
This is the part you can't get from a textbook, and it's where models are genuinely strong: they've seen an enormous number of people get the same thing wrong in the same way. Naming the wrong intuition before you form it saves the week you'd otherwise spend holding it.
The instruction to teach a third of a too-large topic rather than skim all of it is the difference between an hour that leaves you able to do something and an hour that leaves you able to recognise vocabulary.
Produces the orientation doc a new engineer actually needs: the three things that will confuse them, the load-bearing files, and the conventions nobody wrote down.