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.
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."
1. DOCX generation → professional reports in seconds 2. XLSX automation → data analysis without Excel open 3. PPTX creation → entire presentations from outlines 4. PDF processing → merge, split, extract, OCR 5. Web scraping → structured data from any site
Plus: API wrappers, file converters, research tools.