The 8-week curriculum
in full detail.

Four real assets side by side — S&P 500, 10-yr Treasury, a Brooklyn rental, gold. Which made the most 1995–2025? Which lost least in 2008? Which protected you in 2022? Reveal: no single winner.

What an asset actually is (a claim on future cash flows). The four asset classes. The three returns every investor earns: income, growth, inflation protection. Risk premium intuition. Vocabulary fluency: ETF vs. mutual fund vs. index, expense ratio, yield, duration.

Each fellow picks one real ticker (VTI, BND, VNQ, GLD, or a single stock). Prompt AI for a 1-paragraph "what this is, who it's for, what kills it." Fact-check three claims against the actual prospectus or 10-K. Document what AI got wrong.

Two fellows present. Mentor asks: "What did AI tell you that the prospectus didn't say?"

100% S&P vs. 60/30/10 (stocks/bonds/gold) across 2000, 2008, 2020, 2022. Same average return. Wildly different worst years. Which would let you sleep?

Diversification is uncorrelated assets, not many stocks. Correlation intuition without the math. The 3-asset projection model: weights × expected return − fees − inflation = real return. Rebalancing as forced discipline. Regime thinking: base, recession, stagflation, melt-up.

Build your own 3-asset portfolio in a spreadsheet. Use AI to populate historical returns. Run it through three stress regimes: 2008 (−37% equities), 2022 (−13% bonds + −18% stocks), synthetic stagflation (9% inflation, flat equities). Document the worst drawdown.

"Defend your weights. Why 60/30/10 and not 80/20?" Mentor pushes one weakness in writing within 48 hrs.

Long-Term Capital Management in 8 minutes. Two Nobel laureates. $4.6B lost in 4 months. Why? Not bad math — bad behaviour under stress.

Risk ≠ volatility. Risk = permanent loss of capital. Six biases: anchoring, recency, confirmation, loss aversion, overconfidence, herding — with one famous blow-up per bias (LTCM, GameStop, Archegos, FTX retail, dot-com, 2008). Sequence-of-returns risk. Pre-commitment devices: IPS, automatic rebalancing, "don't open the app" rules.

Each fellow writes a 1-page personal Investment Policy Statement (allocation, rebalancing rule, "I will sell if…" rule). Then prompts AI to attack it: "You are a behavioural economist. Find the three biases hiding in this IPS." Revise.

Mentor plays "2008 — your portfolio is down 40%, your friend just sold everything." What does your IPS say?

Two anonymised 10-Ks. One is Costco, one is a near-bankrupt retailer. No names. Which would you lend $10M to? Most fellows get it right in 90 seconds — by reading cash flow from operations, not revenue.

The 3 statements in plain English: income = story, balance sheet = snapshot, cash flow = truth. The 5 numbers that actually matter: revenue growth, gross margin, operating margin, free cash flow, debt/equity. Why earnings lie and cash doesn't. Pattern library: a healthy SaaS, a healthy retailer, a healthy bank, a dying business.

Pick a real public company (NVDA, COST, SBUX, LULU, DUOL). Pull the latest 10-K. Use AI to extract the 5 numbers — then verify each one against the actual filing. Find at least one place where AI rounded, paraphrased, or invented.

"Show me the page number for your gross margin claim." No page = no claim.

Airbnb 2009 vs. a generic 2024 AI wrapper. Both raised money. Only one had a real business. The difference is unit economics.

BMC, but only the 4 boxes that matter for a teenager's idea: customer, value prop, channel, revenue model. Pricing as strategy: cost-plus vs. value-based vs. tiered. Unit economics in one equation: LTV − CAC > 0 and payback < 12 months. The four ways small businesses die: no demand, no margin, no cash, no focus.

Draft a 1-page venture (real or hypothetical — tutoring, dog-walking, an app, an Etsy shop). Build a unit economics model: CAC, ASP, gross margin, payback. Stress-test: what if CAC doubles? Use AI to generate 5 counter-arguments. Keep the strongest two.

"What CAC would kill this business? Show the math."

Two cold emails — one AI-drafted in 4 seconds, one human-rewritten in 4 minutes. The human one converts at 12×. Why?

GTM in one sentence: how a stranger becomes a paying customer. The 5 channels (and the only 2 that work for a teenager): word-of-mouth, content, paid, partnerships, outbound. Message-market fit before product-market fit. The 5 hardest objections every founder hears, and how to answer each in <30 seconds.

Take W5's 1-pager. Use AI to draft three message variants for three channels, then role-play as the 5 hardest objection-givers (a sceptical parent, a competitor, a VC, an unhappy customer, a journalist). Write a tight, sourced answer to each.

Mentor plays the hardest objection live. Unscripted.

60 seconds of a real pig-butchering script. Then a real IRS phishing voicemail. Then a Terms of Service clause that signs away a user's photos. All three are legal-adjacent or outright crimes that target people exactly your age within 3 years.

The 7 scam archetypes that hit Gen Z first — pig-butchering, fake job, fake internship, romance, crypto rug, fake refund, deepfake-of-a-friend — with the one tell that breaks each. Taxes in 30 minutes: W-2 vs. 1099, marginal vs. effective, the only 4 numbers on a paycheck, why a Roth IRA at 17 is the best trade of your life. Contracts you'll sign by 22: lease, employment offer, NDA, ToS, student loan.

Pick one: (a) build a scam-detector checklist for one archetype with 5 real examples, (b) reverse-engineer a real paycheck and find the tax savings, or (c) redline a real lease or ToS and flag the 3 worst clauses. AI drafts; fellow verifies every claim against the source document.

"Defend one clause you flagged. Why is it predatory and not just standard?"