Holy shit…Professors at NYU, Stanford, and Case Western stopped building courses by hand.
They're using NotebookLM to do it in hours and one just called it the biggest shift in academic research in 20 years.
Here's the exact workflow they shared publicly:
Step 1 - The source upload strategy that changes everything.
Most people upload one or two documents.
Professors building full courses upload their entire reading list at once.
Up to 50 sources per notebook on the free tier.
500,000 words per source.
PDFs, Google Docs, URLs, YouTube lectures, audio files, images with OCR.
One notebook = one course unit.
The workflow Stanford faculty documented publicly:
→ Upload all assigned readings for a unit
→ Upload the course syllabus
→ Upload previous years' exam questions
→ Upload any relevant primary sources
NotebookLM now has a 1-million token context window.
It holds the entire unit in its head simultaneously and reasons across all of it at once.
No AI on the market does this grounded in YOUR specific sources.
Step 2 - The curriculum mapping prompt.
Once sources are uploaded, this is the first prompt professors run:
"Based on all uploaded materials, create a complete curriculum map for this unit. Identify the 5 core concepts students must understand. For each concept, list: the source that introduces it, the source that deepens it, and the source that challenges or complicates it. Then suggest a logical teaching sequence."
What comes back is a structured roadmap of the entire unit with every claim cited back to the exact uploaded source.
NYU's Assistant Dean used a version of this to identify course equivalencies across an entire revised curriculum for student advising.
What used to require weeks of manual cross-referencing across departments happened in one session.
Step 3 - Generating the lesson plan stack.
After the curriculum map, professors run this:
"Using the uploaded syllabus and course materials, create a detailed day-by-day lesson plan for [topic]. Each class session should include: a learning objective, the key concepts to cover, one real-world example or analogy, a discussion question, and an estimated time breakdown for a 60-minute class."
NotebookLM generates the full plan grounded strictly in the materials uploaded.
No hallucinated citations. No invented examples.
Everything it produces traces back to a source you can click and verify inline.
Arizona State University faculty documented using exactly this approach to organize scholarly articles and identify themes across disciplines work that previously took entire research semesters.
Step 4 - The student-facing materials workflow.
This is where NotebookLM saves professors the most time per week.
After building the lesson plan, they generate the entire student-facing package from the same notebook:
→ Study guides with main ideas, critical arguments, and supporting evidence - one click in Studio
→ Reading comprehension questions based on exact course materials
→ Flashcards for key concepts and terminology - auto-generated, cited to source
→ Practice quizzes - multiple choice or short answer, with an answer key that cites back to the reading
Northeastern University lecturers documented having students create their OWN custom study aids from course readings using this exact workflow.
Students stop asking "what do I need to know?"
The notebook tells them in their own preferred format from the exact materials the professor assigned.
Step 5 - The Audio Overview for student prep.
This is the feature that went viral among students once professors started sharing it.
Before every class, professors generate an Audio Overview of that week's readings.
Students get a 10-15 minute podcast-style summary two AI hosts explaining the material, making connections, using analogies — before they've read a single page.
Case Western Reserve University faculty documented using this specifically to help students approach dense readings they'd otherwise avoid entirely.
The result: students show up to class having already encountered the key ideas.
Class time shifts from "let me explain what the reading said" to actual discussion, application, and debate.
That's the pedagogical shift professors are calling transformational.
Not the tool itself. What the tool makes possible in the room.
Step 6 - The Deep Research upgrade for literature reviews.
This is the feature that changed things for research faculty specifically.
Deep Research available on NotebookLM lets the tool autonomously search the web, build a bibliography, and compile a fully cited research report.
One Pitt researcher documented cutting literature review prep time by 70% using it.
The workflow:
→ Upload your existing sources and research question
→ Run Deep Research
→ NotebookLM plans its own web searches, identifies gaps in your current sources, pulls new papers, and synthesizes everything into a cited report
Walter Isaacson used NotebookLM to analyze Marie Curie's journals for his book.
Primary historical research documents not just student notes.
When a Pulitzer Prize-winning biographer is using your research workflow, that's a signal worth paying attention to.
Step 7 - Google Classroom integration for sharing with students.
This is the 2026 update that closed the loop for faculty.
You can now create a NotebookLM notebook directly from Google Classroom.
One click pulls in all the resources already assigned to students.
No manual re-uploading. No rebuilding the curriculum in a new tool.
Assign notebooks to students as "View Only" - they get the full AI-assisted experience grounded in exactly the materials you assigned.
Students can query the notebook, generate their own study guides, run their own audio overviews, create flashcards all from the same sources the professor built the course on.
The professor sets the knowledge base once.
The students interact with it in whatever format helps them learn best.
That's personalized education at scale without any extra work from the professor after the initial setup.
The full professor workflow - save this.
Here's the complete stack documented by faculty across NYU, Stanford, Case Western, Arizona State, and Northeastern:
Step 1 → Upload entire reading list + syllabus + past exams into one notebook per unit
Step 2 → Run curriculum mapping prompt get the full teaching sequence with citations
Step 3 → Generate day-by-day lesson plans for every class session
Step 4 → Generate student-facing package study guides, flashcards, quizzes in one click
Step 5 → Create Audio Overviews for student pre-reading before every class
Step 6 → Use Deep Research to run literature reviews and identify source gaps
Step 7 → Share notebooks via Google Classroom students interact with the same source base
What used to take a full summer now takes a week.
What used to take a week now takes an afternoon.
The professors who figure this out first aren't working less.
They're working on the parts that actually require a human in the room.
The university system has been running on the same curriculum-building process for decades.
Upload readings. Write lectures. Build assessments. Repeat every semester from scratch.
NotebookLM didn't change what professors teach.
It changed how long the invisible work takes.
The literature review that took 3 months.
The lesson plan stack that took 2 weeks.
The study materials that took a weekend.
All of it is still the professor's intellectual work.
NotebookLM just removed the part where you manually cross-reference 50 PDFs at 11pm.
The institutions moving fastest on this aren't replacing faculty.
Whenever I'm stuck on a business decision for more than 3 days, I run these 10 Claude prompts and get more clarity in 15 minutes than 3 weeks of overthinking ever gave me.
Copy and paste them into any chat:
1. THE REAL QUESTION FINDER
Before you solve anything, you need to know what you're actually solving.
Most business decisions feel hard because you're asking the wrong question.
Paste this:
"I've been stuck on this decision for [X] days: [describe the decision]. Before giving me advice, reframe this for me.
Tell me:
1) What the surface-level question appears to be, 2) What the actual underlying question I'm wrestling with really is, 3) What fear or assumption is driving my indecision, 4) What the real decision would look like if I stripped away the noise."
You'll realize you've been debating the wrong thing the entire time.
2. THE 10/10/10 RULE
Suzy Welch built her entire career on this framework.
Claude applies it in 30 seconds.
"Run my current decision through the 10/10/10 test:
1) How will I feel about this decision 10 minutes from now, 2) How will I feel about it 10 months from now, 3) How will I feel about it 10 years from now, 4) Based on all three time horizons, tell me which option creates the most long-term value and which one I'm avoiding out of short-term discomfort."
Most bad decisions look smart at 10 minutes and stupid at 10 years.
Perplexity Computer has been quietly shipping features that make every other AI tool feel like a chatbot.
Most people don't even know these exist.
Here are 10 things it can do right now that will change how you work ↓
𝟏. Scheduled Tasks That Run While You Sleep
Computer doesn't just answer questions — it runs jobs on a schedule.
→ "Every Monday at 7am, audit my website SEO and email me the report"
→ "Every morning, pull competitor pricing changes and flag anything new"
→ "Every Friday at 5pm, compile my team's weekly metrics from Slack and Sheets"
You set it once. It runs forever. No reminders. No manual work.
You wake up and the work is already done.
𝟐. 20+ AI Models. Zero Switching.
Computer doesn't use one model. It uses over 20.
And here's the part nobody realizes — it picks the best model for each step automatically.
→ Research step? Routes to the best model for web retrieval.
→ Writing step? Routes to the best model for long-form.
→ Analysis step? Routes to the best model for reasoning.
→ Image generation? Routes to the best visual model.
One prompt. Multiple models working behind the scenes.
You never switch tabs, accounts, or subscriptions again.
BREAKING: AI can now analyze any stock like a Wall Street analyst (for free).
Here are 10 insane Claude prompts that replace $2,000/month Bloomberg terminals:(Save for later)
1. The Goldman Sachs Stock Screener
"You are a senior equity analyst at Goldman Sachs with 20 years of experience screening stocks for high-net-worth clients.
I need a complete stock screening framework for my investment goals.
Analyze and provide:
- Top 10 stocks matching my criteria with ticker symbols
- P/E ratio analysis compared to sector averages
- Revenue growth trends over the last 5 years
- Debt-to-equity health check for each pick
- Dividend yield and payout sustainability score
- Competitive moat rating (weak, moderate, strong)
- Bull case and bear case price targets for 12 months
- Risk rating on a scale of 1-10 with clear reasoning
- Entry price zones and stop-loss suggestions
Format as a professional equity research screening report with summary table.
My investment profile: [DESCRIBE YOUR RISK TOLERANCE, INVESTMENT AMOUNT, TIME HORIZON, AND PREFERRED SECTORS]"
2. The Morgan Stanley DCF Valuation
"You are a VP-level investment banker at Morgan Stanley who builds valuation models for Fortune 500 M&A deals.
I need a full discounted cash flow analysis for a specific stock.
Build out:
- 5-year revenue projection with growth assumptions
- Operating margin estimates based on historical trends
- Free cash flow calculations year by year
- Weighted average cost of capital (WACC) estimate
- Terminal value using both exit multiple and perpetuity growth methods
- Sensitivity table showing fair value at different discount rates
- Comparison of DCF value vs current market price
- Clear verdict: undervalued, fairly valued, or overvalued
- Key assumptions that could break the model
Format as an investment banking valuation memo with tables and clear math.
The stock I want valued: [ENTER TICKER SYMBOL AND COMPANY NAME]"
BREAKING: AI can now build you a business in 24 hours.
Here are 16 insane Grok 4 prompts to turn any idea into income in 2026: (Save for later)
1/ THE BUSINESS IDEA GENERATOR
"I have $500 and 10 hours per week. Analyze current market gaps, my skills in [your skills], and generate 5 business ideas I could launch this month. Include startup costs, timeline, and first revenue projections."
This prompt is stupid powerful.
2/ THE COMPETITOR DESTROYER
"Act as a market analyst. Research [competitor name] and identify 3 weaknesses in their offering. Then create a differentiation strategy I could execute with limited budget to capture their customers."