LLenny's Podcast
← All frameworks
ProductivityJulie Zhuo (Facebook VP, Sundial CEO, The Making of a Manage

AI as Your Personalized Just-In-Time Tutor

Feed AI a curriculum tuned to how you learn, then prove understanding by teaching it back.

Difficulty
Easy
Time to result
~days to results
Steps
4
Confidence
85%

Zhuo argues the most underutilized use of AI is accelerating your own learning, not just autocompleting code. Because you no longer have to consume a human colleague's time to learn a skill, you can hand an AI a curriculum, ask it to personalize the program to your learning style, and then test your understanding by explaining concepts back in your own words and letting it critique you. AI is often a better teacher because it personalizes and is interactive.

Origin

Julie Zhuo's practice at Sundial, where engineers use ChatGPT to learn data analysis and the 'why/when' behind techniques.

Core principles

  • 01The bottleneck used to be a human's time to teach you; AI removes it
  • 02AI personalizes to how each person learns — examples and analogies for some, terse for others
  • 03AI teachers are especially good for general-purpose 'why' and 'when' questions with abundant public knowledge
  • 04Interactivity lets you test real understanding, not just consume explanations
  • 05A focused 30-60 minute session can teach what used to take far longer

How to run it

  1. 1

    Find a curriculum and hand it to the AI

    Locate an existing course outline (e.g., a 12-week curriculum) and feed it into ChatGPT.

    Pro tip Ask it to customize the program to how you like to learn.

  2. 2

    Specify your learning style

    Tell it how you learn best — e.g., 'I need examples, analogies, explain like I'm five' — since learners genuinely differ.

    Pro tip Different teammates need opposite styles; the point is per-person personalization.

  3. 3

    Ask the why and when, not just the how

    Use AI for the general-purpose context questions — why or when a technique is useful — that you'd otherwise have to ask a busy human.

    Pro tip These broadly-documented questions are exactly where AI gives a richer, deeper answer.

  4. 4

    Test yourself by teaching it back

    After reading, explain the concept back in your own words and analogy, and let the AI critique whether you got it right.

    Pro tip It nicely says 'that's close' before telling you if you were completely wrong — use its critique to find gaps.

In the wild

The engineer learning the 'why' of root-cause analysis

An engineer at Sundial understood how to run root-cause analysis algorithms but not why or when they'd matter in a company, having never been a PM or executive. Zhuo notes this general-purpose question is 'the perfect type of question' to take to ChatGPT for a better, deeper answer than asking a colleague.

The engineer learns context faster and deeper without consuming a coworker's time.

Teach-back to verify understanding

Zhuo reads something, then tells ChatGPT 'so this means...' with her own analogy, and it critiques whether she's right — 'we can really test whether we really understand the concept by trying to retell it back in our own way.'

Genuine comprehension checks rather than passive consumption.

Common mistakes

Forgetting AI can teach, not just do

People use AI to autocomplete and produce output but forget it can accelerate their own education, missing a 30-60 minute path to learning skills far faster than before.

Passively consuming explanations

Reading AI output without teaching it back leaves you unsure you actually understood; skipping the interactive critique step wastes AI's biggest learning advantage.

Is it for you?

Best for

Builders picking up an unfamiliar discipline (e.g., engineers learning data analysis) who want to self-teach fast

Not ideal for

Highly proprietary or novel domains with little public knowledge for the model to draw on

From the transcript

I'll find a curriculum online

23:00

I'll just feed it into chat PT and I'll say help me customize a program for me

23:30

these days, you know, TAGBT or these other AI tools are better teachers

23:00

try to explain back what I heard

26:00

we can really test whether we really understand the concept by trying to retell it back in our own way

26:30

From the episode

From managing people to managing AI: The leadership skills everyone needs now

Julie Zhuo (Facebook VP, Sundial CEO, The Making of a Manage