The Semantic Tree Learning Prompt
Teach yourself anything from the roots up using an LLM
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 78%
George Mack combines Elon Musk's 'knowledge as a semantic tree' idea with an LLM to teach himself any subject. Rather than jumping straight to a specialist podcast on, say, the heart, you treat the field as a tree: roots, trunk, primary branches, secondary branches, then leaves. You prompt Claude or ChatGPT with the Musk framing and say 'teach me about X, but start with the roots and work all the way up, and don't move to the next layer until I say I understand.' You keep confirming understanding layer by layer, then transfer the structure into mind-mapping software so you build the tree in your own head. The premise is that knowledge is the only real bottleneck, and with the alphabet, numeracy, and an LLM you can reach any concept.
Origin
George Mack's 2024 development of his prior year's 'treat me like an idiot and start at zero' prompt, merged with Elon Musk's semantic-tree learning quote.
Core principles
- 01Knowledge is a tree: roots and trunk before branches and leaves
- 02You can't retain leaves you hang on nothing
- 03The only true bottleneck to most goals is knowledge
- 04Force sequential mastery: don't advance until you actually understand
How to run it
- 1
Choose the postponed topic
Pick something you've always wanted to learn but keep kicking down the road because you don't know where to begin.
Pro tip George started with longevity for exactly this reason.
- 2
Prime with the tree framing
Give the LLM the Musk semantic-tree quote so it structures the material as roots up to leaves.
- 3
Demand root-first sequencing
Instruct it: teach me about X, start with the roots and work all the way up, and don't move to the next layer until I say I understand.
Pro tip The 'until I say I understand' clause is what forces genuine retention.
Watch out Don't let yourself wave through a layer you haven't actually grasped.
- 4
Confirm layer by layer
Only advance once you can hold the current layer; keep climbing trunk, branches, then leaves.
- 5
Externalise into a mind map
Move the structure into mind-mapping software and build the tree yourself so the interweaving parts live in memory.
Pro tip This scaffold is what future facts hang on; it prevents raw-data memorisation.
In the wild
George had always wanted to understand longevity but kept deferring it because he didn't know where to start. Using the prompt, he began at the roots of the field and worked upward, only advancing when he'd grasped each layer, turning an intimidating topic into a structured, learnable tree.
→ A topic he'd procrastinated on for years became something he could start and actually retain in a day.
Common mistakes
Jumping straight to the leaves
Starting with an advanced specialist interview gives you no roots, so the information never connects and doesn't stick.
Brute-forcing raw data
Memorising disconnected facts with no skeleton to hang them on is rough going and rarely becomes real understanding.
Is it for you?
Best for
Self-directed learners tackling a brand-new domain they've been intimidated by.
Not ideal for
Quick factual lookups or topics where you already have a strong foundation.
From the transcript
“you say the specific Elon quote, and you say, Teach me about X, but start with the roots and then work all the way up…”
“you have to have like a framework or a skeleton to be able to hang concepts on. Otherwise you are just learning raw data and…”
From the episode
Christmas Special - Life Hacks, Biggest Lessons & Best Resolutions - #881