What Buffett Would Do

3 worked examples

Each example shows the full verdict + 3 Check First + Next Action format, on a real investment thesis. No scorecards. No buy / sell.

Verdict mix: Research Now ×1 · Watch ×1 · Wait for Evidence ×1

These are prompt demonstrations, not stock recommendations. Different data, timing, assumptions, and user position context can produce different verdicts.

1. Progressive (PGR) — insurance underwriting discipline and float

Verdict: Research Now

Why this verdict, not the adjacent ones. Underwriting discipline and float economics are visible across cycles, and the current price is not so far out of band that the test fails before the deep-dive starts.

Check First (exactly 3)

  1. Is the current combined ratio inside Progressive's long-term target band, and is the trend improving or stable?
  2. What does today's price imply about forward combined ratio and premium growth?
  3. How is float yield trending vs. the long-term risk-free rate, and is the spread to risk-free + risk premium adequate?

Next Action

Open as Research Now. Run a deep-dive prompt focused on (a) underwriting trend, (b) float economics vs. peer set, (c) capital allocation between buybacks and dividends.

2. Nvidia (NVDA) — exceptional business, monitoring required on multiple fronts

Verdict: Watch

Why this verdict, not the adjacent ones. Real quality, but valuation, custom silicon (TPUs / ASICs), export-control scope, and a 5+ year underwriting path that's not yet visible together rule out immediate deep research.

Check First (exactly 3)

  1. Is the FCF / earnings yield spread to the long-term risk-free yield + risk premium wide enough to compensate for the GPU architecture cycle, custom silicon, and export-control risk?
  2. What is hyperscaler capex commentary saying about the durability of the training cycle, and is the implied growth rate in the current price plausible against that commentary?
  3. What evidence would actually change the verdict to Pass — a multiple that reflects ordinary expectations, or a clearer 5+ year underwriting path through custom silicon and export controls?

Next Action

Stay on Watch. Re-triage when any of (a) a credible 2nd-source AI accelerator ecosystem gains real workloads, (b) hyperscaler capex commentary flattens for two consecutive quarters, (c) export-control scope materially changes the addressable market, or (d) the multiple compresses enough that the FCF-yield spread to risk-free + risk premium becomes clearly attractive.

3. Costco (COST) — great business, may still be expensive

Verdict: Wait for Evidence

Why this verdict, not the adjacent ones. High-quality business, but the multiple is again near the upper end of its 10-year range and the price already assumes a long tail of strong comp and member-fee growth — proof point not yet visible.

Check First (exactly 3)

  1. Is the current EV/EBITDA inside or outside Costco's 10-year band?
  2. What FCF / earnings yield is the market demanding, and how wide is the spread to the long-term risk-free yield plus a risk premium?
  3. What does today's price imply about forward comp growth and membership-fee growth?

Next Action

Stay at Wait for Evidence until either (a) the multiple compresses enough that the FCF-yield spread to risk-free + risk premium becomes clearly attractive, or (b) earnings confirm the implied growth is being delivered. Re-triage at Q1 / Q2 2026 earnings.

Save this verdict for future review

Want to save your own research-priority verdict for future review? Save it here.

Save this verdict for future review

Public tracking cases

Live, time-stamped case studies WBWD tracks in public. Different from the worked examples above: these are not illustrative, they are ongoing and updated when material events occur.

Momenta (港股 AI IPO) — three-layer view, live review schedule, public tracking log

Public tracking case

Watch — three layers kept separate; tracking log updated as milestones fire.

Last reviewed 2026-07-07 · next planned review 2026-07-08 (Listing day).

View public case →

Momenta is WBWD's first public tracking case. New cases are added when material events warrant, not on a calendar.