Millie Marconi Profile picture
Apr 10 8 tweets 5 min read Read on X
Claude can now teach you how to think using the exact method Richard Feynman used at Caltech for 40 years.

Most people use Claude to get answers.

These 5 prompts use it to rewire how fast you learn anything 👇
1/ The Confusion Locator

Feynman said the first step in understanding anything is being honest about what you actually don't understand versus what you just can't explain.

Most people confuse familiarity with understanding.

They've heard a term enough times that it feels known. But the moment they try to explain it, the gaps appear.

"I think I understand [concept] but I want to test that. Ask me to explain it to you as if you're a curious 12-year-old who has never heard of it. After I explain it, tell me: where did my explanation break down or get vague? Where did I use words that assume prior knowledge the 12-year-old wouldn't have? Where did I skip a logical step that I assumed was obvious? Give me a precise list of every gap you found. Those gaps are exactly what I don't actually understand yet."

The gaps this prompt surfaces are more valuable than anything you'd learn from re-reading the source material.

Because they're your specific gaps.

Not the gaps of the average reader.
2/ The First Principle Finder

Feynman never started from the middle of a subject.

He always started from what was actually true at the most fundamental level the irreducible facts that everything else in the field was built on top of.

His first Caltech lecture didn't start with Newton's laws.

It started with the atomic hypothesis. The one idea that if everything else was lost to science would contain the most information in the fewest words.

"I am trying to understand [subject]. Don't teach me the standard curriculum. First: what is the single most fundamental true statement about this subject? The one idea that if I understood it completely would make every other concept in this field easier to learn? Build my understanding from that single statement outward, adding only one layer of complexity at a time, and stopping to check whether each layer is actually clear before adding the next."

The student who starts from first principles always overtakes the one who started from the textbook.

Because foundations compound.

Surface knowledge doesn't.
3/ The Analogy Builder

Feynman's most powerful teaching tool wasn't equations.

It was analogies.

He could make quantum electrodynamics feel intuitive to a non-physicist because he had spent years finding the everyday physical experience that captured the essence of what the mathematics was actually describing.

The analogy isn't a simplification. It's a bridge.

"I am struggling to intuitively understand [concept]. Don't explain it using the standard technical language. Instead: find the everyday physical experience, familiar object, or common situation that captures the essential behavior of this concept. The analogy should work without any prior knowledge of the field. Test your analogy by explaining what it gets right about [concept] and where it breaks down — because every analogy breaks somewhere and knowing where is part of the understanding."

The moment an analogy clicks is the moment abstract knowledge becomes usable knowledge.

Feynman knew this.

Most textbooks never learned it.
4/ The Teaching Test

Feynman's most famous learning technique is also his most misunderstood.

Everyone knows "teach it to a child."

What people miss is the second half.

When the explanation breaks — and it always breaks somewhere — you don't go back to the book and re-read it.

You go back to the fundamental principles and rebuild the explanation from scratch.

The rebuilding is the learning.

"I want to use the Feynman technique to learn [concept]. First: let me explain it to you in the simplest language I can. [Write your explanation here.] Now: tell me exactly where my explanation broke down, got vague, or relied on terminology I didn't actually define. For each breakdown point: don't just tell me what's wrong. Give me the simplest accurate explanation of that specific piece so I can rebuild my understanding from the ground up. I want to iterate this until I can explain the full concept in plain language with zero gaps."

Iterate this three times on the same concept.

By the third iteration you'll understand it better than most people who've studied it formally.
5/ The Integrated Understanding Test

Feynman's final test for genuine understanding was performance under pressure.

Not a quiz. Not a recall exercise.

A novel problem that required combining everything learned into something that had never been explicitly taught.

Because real understanding isn't knowing answers.

It's being able to generate answers to questions you've never seen before.

"I have been learning [subject] using Feynman's method across these concepts: [list what you've covered]. Now give me a novel problem something that requires combining at least 3 of these concepts simultaneously to solve, that I have definitely never seen before, and that has no single right answer but does have better and worse answers based on how deeply I understand the material. After I attempt it: tell me what my answer reveals about where my understanding is genuine versus where it is still surface level. What would Feynman's answer look like and what does the difference between his answer and mine reveal about what I still need to work on?"

This is the test that separates studied knowledge from owned knowledge.

Feynman gave versions of this to his Caltech freshmen.

It's the hardest test in any subject.

It's also the only one that matters.
Feynman spent 40 years at Caltech doing one thing.

Making hard things simple.

Not dumbing them down. Making them genuinely clear.

He believed that if you couldn't do that, you didn't understand the thing yet.

No exceptions.

Not for Nobel laureates. Not for graduate students. Not for anyone.

These 10 prompts run his standard on whatever you're learning.

Not to make you feel smart about a subject.

To find out whether you actually understand it.

The difference between those two things is the whole point.

Save this thread. Pick one concept you think you understand.

Run prompt 1 first.

Find out where the gaps actually are.

That's where the real learning starts.
AI makes content creation faster than ever, but it also makes guessing riskier than ever.

If you want to know what your audience will react to before you post, TestFeed gives you instant feedback from AI personas that think like your real users.

It’s the missing step between ideas and impact. Join the waitlist and stop publishing blind.

testfeed.ai

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Millie Marconi

Millie Marconi Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @MillieMarconnni

Apr 7
After using Perplexity for 2 years, it has become the only research tool I open before anything else

But only because I stopped using it like Google

Here's the 8 prompt system that turns a topic I know nothing about into a 30-page research brief in under an hour. Image
1. Start with ignorance mapping

Prompt:

"I know almost nothing about [topic]. Before I start researching, what are the 10 things I most need to understand first? Order them from foundational to advanced. Tell me which ones most people get wrong."

This is not a search. It's a curriculum.

Most people skip this and spend 3 hours reading the wrong things in the wrong order.
2. Find the real debate

Prompt:

"What do experts in [topic] genuinely disagree about right now? Not surface-level stuff. The real fault lines where smart people are on opposite sides and both have evidence."

Google gives you the consensus.

This prompt gives you where the field is actually alive.
Read 11 tweets
Apr 2
After 6 months of using NotebookLM, I can say it's the research tool that has revolutionized my workflow the most.

But only because I learned these 10 prompts.

Here's the complete system that turns 200 pages into clear answers in under an hour: Image
1. The Source Onboarding Prompt

Before you do anything else, run this the moment you upload your documents.

Paste this into NotebookLM:

"You now have access to [X] sources I've uploaded. Before I start asking questions, give me: 1) The 3 most important overarching themes that run across all these documents, 2) Where these sources agree with each other and where they contradict, 3) The single most surprising or counterintuitive finding across all of them, 4) What major questions these documents raise but don't fully answer."

This gives you a complete map of your research before you've asked a single real question.

Most people skip this. Don't.
2. The Contradiction Hunter

This is where NotebookLM becomes genuinely dangerous in the best way.

Use this:

"Go through all my uploaded sources and find every place where two or more sources disagree, contradict each other, or come to different conclusions on the same topic. For each contradiction: 1) Quote the specific conflicting claims, 2) Identify which source each claim comes from, 3) Give me your assessment of which position has stronger supporting evidence, 4) Flag this as something I need to investigate further."

The best research lives in the gaps between sources.

This prompt finds every single gap.
Read 12 tweets
Mar 27
🚨 BREAKING: HuggingFace just dropped their complete AI engineering playbook to the public.

They released 12 courses that were internal-only until this week.

This covers LLMs, Robotics, and MCP, which is the exact tech stack behind Llama, Mistral, and every major open model.

This level of training won't stay free forever.

Here's what you need to grab right now 👇Image
1/ LLM Course

This is where you start if you're serious about AI.

Not "what is an LLM" baby content.

Actual hands-on training fine-tuning models, building pipelines, working with the full HuggingFace ecosystem from transformers to tokenizers to deployment.

Most $2,000 bootcamps teach less than this.

It's free. There's no excuse anymore.
huggingface.co/learn/llm-cour…
2/ Agents Course

This is the most important course on this list for 2026.

Everyone's talking about AI agents. Almost nobody understands how they actually work under the hood.

This course teaches you to build and deploy your own tool use, memory, multi-step reasoning, the full stack.

Not theory. Not demos. Actual deployable agents.

If you're building anything with AI this year, start here after the LLM course.
huggingface.co/learn/agents-c…
Read 14 tweets
Mar 17
This is wild.

MIT researchers proved you can make ChatGPT review and improve its own work using self-critique prompting.

I've been using it for 3 months and it completely changed my results.

Here are 8 prompts that actually work:
The paper is called Self-Refine.

The finding is embarrassingly simple:

LLMs don't give you their best answer first.

They give you a first draft.

The difference between a mediocre answer and a great one?

Asking it to review its own work. Image
Prompt 1: The Weakness Hunt

After any answer, send this:

"List the 3 biggest weaknesses in the response you just gave me. Be specific and brutal. Then rewrite it fixing those weaknesses."

Works on emails, strategy docs, essays anything.
Read 12 tweets
Mar 14
This is wild...I fed Claude my Amazon listing and asked it to think like a top 1% seller.

What it did in 8 minutes would've cost me $3,000 at an agency.

Here are 12 prompts stealing their entire playbook: Image
1. The “A+ Listing Builder” Prompt

Prompt:

"Act as an Amazon conversion copywriter. Write a complete product listing for {product}. Include an SEO-optimized title, 5 high-converting bullet points, a persuasive product description, and A+ content sections. Focus on benefits, not just features."

This creates conversion-focused listings.
2. The “Competitor Keyword Intelligence” Prompt

Prompt:

"Analyze the top 10 Amazon listings for {product category}. Extract the most common keywords used in titles, bullet points, and descriptions. Identify the keywords driving visibility."

This reveals what competitors rank for.
Read 14 tweets
Mar 11
After 2 years of using Claude, I built a second brain that remembers my goals, patterns, and blind spots.

It's changed everything about how I work and think.

Here's how to build it (save this): 👇 Image
Step 1: Build your Master Context File.

Open a doc. Write:

→ Your 3 core goals this year (specific, not vague)
→ Your top 3 constraints (time, money, skills)
→ Your default blindspots (what do you always get wrong?)
→ Your decision-making style (gut? data? both?)

Paste this at the start of every Claude session.
This one habit changes everything.

Instead of "help me write an email," you're saying:

"Given that my goal is X, my constraint is Y, and I tend to overlook Z help me write this email."

The output quality isn't even comparable.

Context is leverage. Most people throw it away every chat.
Read 11 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(