Actionable insights — Amazon Blocks Meta: The Agentic Wars Have Started

2:53 1. Screen for the coiled setup, then wait for the story change

1. Screen for three fundamentals together: a low forward P/E, fast earnings growth, and a wide, defensible moat (ask whether AI or lawsuits can realistically disrupt it).
2. Add the fourth condition that makes it coiled: negative sentiment. The media and analysts are bearish, and the stock trades as if the bad story is permanent.
3. Clear the overhangs one by one (settled litigation, resolved regulatory cases). Each one removed "makes the slate a little more clean."
4. Own it on the fundamentals alone ("enough reason in and of itself to buy the stock"). Then expect the move when a story change arrives, even with no change to EPS estimates.

• Analyst target hikes that follow the price rather than estimate revisions (Wells Fargo $640→$796). That is a sign the story, not the numbers, is driving the move. Also watch for the catalyst event (Meta Connect) where the story either deepens or fades.

12:25 2. Test a company's stated reason with a control case

1. Write down the official reason for a strategic move (a block, a price change, an exit).
2. Find a case where that reason does not apply and see whether the behaviour changes. If the risk is really about credentials, a request without credentials should get through.
3. If the behaviour stays the same without the stated cause, treat the stated reason as cover and look for the economic one (next insight).

• A policy change that closes the gap, such as allowing logged-out agents or verified agents only. That would make the stated reason more credible and show where the negotiation is heading.

13:11 3. Find the revenue line a new technology devalues

1. For an incumbent facing a new interface, list revenue that depends on human attention at a fixed spot: sponsored listings, top-of-page placement, in-app promotions.
2. Size it, and note its margin. High-margin ad lines hurt most.
3. Ask what happens if a machine does the looking: it scans every result for the best deal, so paid placement loses value and advertisers pay less.
4. Add the second-order losses: the customer touch point, behavioural data, and the incumbent's own competing tool.

• Retail-media growth decelerating as agent traffic rises, and incumbents launching "agent portals" or paid agent access, which would turn the threat into a new fee line.

16:57 4. Read an incumbent's block as a negotiation, and as validation

1. Ask who loses if the block lasts. If both sides lose ("mutually destructive"), it is an opening position, not an end state.
2. Track how the balance shifts with the entrant's scale: worse for the entrant today, worse for the incumbent as the entrant grows.
3. Identify the entrant's outside option, such as a rival willing to open up (Walmart), which raises the incumbent's cost of holding out.
4. Treat the block itself as evidence: incumbents don't block irrelevant products.

• A revenue-share or certified-agent announcement (his base case); other retailers following Amazon's block (a coordinated stance would weaken the thesis); and Muse's usage growth, which sets the timetable.

19:52 5. Grade your own holdings on execution against the natural winner

1. For a new product category, ask which company should win on assets (data, trust, integration, distribution).
2. Check who actually shipped first and how good it is.
3. If a holding had the assets and lost the race, record it as an execution miss, and say so even as a shareholder, without confusing it with a sell signal.

• Google's and OpenAI's agent launches (he predicts an OpenAI counter-launch during Meta Connect) and whether Muse's lead survives them.

21:57 6. Log the bearish TV calls made at the lows, then grade them

1. When a stock is at "very low valuation territory" and "every single negative thing" has been said, note who goes on air to add more.
2. Record the call and the date, then revisit it after a month.
3. Grade the call, not the person: credit the commentator's good calls elsewhere.

• A cluster of bearish media takes arriving after a drawdown and without new facts. That is usually the sentiment low, not new information.

25:11 7. Grade a scary claim by its sourcing chain, plausibility and track record

1. Trace the chain: who observed it, and how many removes is the speaker from them? A "belief" held by an unnamed person, relayed by someone else, is two removes from knowledge.
2. Ask for any independent evidence (company statements, researchers, regulators).
3. Test technical plausibility against how the system actually works.
4. Check the speaker's record on past predictions of the same kind.

• AI-doom headlines that move AI stocks without corroboration, which can be buying opportunities in names like META (his Sep-14 call). Also watch whether such claims feed regulatory proposals, which are the real policy risk.