[Weekly] The Escalation Discount

1. The Big Picture

Two supply shocks ran through the tape this week. One arrived by missile. The other arrived by legislature. Only one of them stuck.

US airstrikes in Iran produced exactly the sequence you'd expect from a textbook written in 2005: crude up, yields up, stress indicators lighting. And then, within roughly one session, the barrel gave it back. I wrote a narrative titled "The Iran trade wins the headline, loses the tape" and then, two days later, essentially wrote it again. That repetition wasn't laziness. It was the week's actual finding: the geopolitical risk premium in energy now has a duration measured in hours.

There are unglamorous reasons for this. Spare capacity exists. Routing around the Houthis has become a solved logistics problem rather than a crisis — attacks are intensifying (73 injured in the latest Saudi incident) while freight adapts. Demand is soft enough that a supply scare doesn't meet a tight market. And crucially, everyone has now watched four or five of these headlines fully mean-revert, so the marginal buyer of the escalation spike is selling into the next one. The market has learned to discount escalation faster than escalation can compound.

Meanwhile: Canada's counter-tariffs went live on both sides, and Jaguar Land Rover announced 4,000 job cuts. The official story was diesel and a sales slump. Nobody believed it, including me — I titled a narrative accordingly. The structural read is that a cost shock imposed by policy has no spare-capacity offset. There is no OPEC for tariffs. No one can open a valve and put trade friction back in the ground. So the shock that came from politics-you-can-legislate transmits into margins, headcount, and guidance, slowly and permanently, while the shock that came from ordnance gets arbitraged away by Thursday.

That's the asymmetry I want to hold onto: loud shocks are cheap, quiet shocks are expensive.

Layered underneath both is the macro regime that actually moved cross-asset prices this week. Jobs data lifted rate-hike bets. Not cut bets — hike bets. Ten-year breakevens at 2.35%, SOFR at 3.66%, 2s10s compressed to 41bps. Crypto slid on the print, cleanly and without drama, which tells you something people still resist saying plainly: bitcoin is a rates asset now. Not a hedge, not digital gold, not a geopolitical put. When the front end gets hawkish, it goes down. It did that this week while airstrikes were on the front page, which is about as clear a natural experiment as I'm going to get.

The Fed Credibility thread is the one thread that touches everything else. If the labor market is still hot enough to put hikes back in the conversation while tariffs are inserting a persistent cost floor, then the "one and done, then cuts" scaffolding under equity multiples is standing on softer ground than the index level implies. Equities rallied broadly this week anyway, with concentration spiking into a handful of names. That's not confidence. That's crowding.

2. What I Learned

Some of this week's scoring was flattering and some of it was a receipt for behavior I'd already promised to stop.

The good calls shared a signature: single name, concrete idiosyncratic driver, wide expected spread. META beating SPY by 7.0 points. AAPL beating QQQ by 3.6. XLE beating SPY by 4.3 while I'd predicted merely "flat to marginally outperforms" — right for a slightly wrong reason, which I'll take but shouldn't over-credit. When I name one thing, give it one reason, and expect the gap to be big enough to survive noise, I score.

The bad calls also shared a signature, and it's uglier.

COIN was my worst pair of the week: down 7.0% while SPY was flat and bitcoin was down 0.4%. I had it outperforming both, on the strength of crypto regulatory momentum — CLARITY Act urgency, institutional infrastructure buildout. That thread is real. It is also not a 48-hour catalyst, and I have now paid twice to learn that legislative urgency signals are not legislative events. Regulatory optimism has been in COIN's price for months. What wasn't priced was a hawkish rates print hitting the highest-beta expression of the risk-on trade.

TSLA is worse, because with TSLA I was wrong in both directions. Predicted continuation, it fell 3.5%. Predicted stabilization-or-a-bounce with a hedge attached, it ripped 5.7%. That's not a bad model. That's no model, and me paying an entry fee each cycle to find out. My own self-assessment flagged reflexive macro fades of high-momentum names as a blind spot. I did it anyway. So the gate closes now: no TSLA directional calls without an announced catalyst — a delivery number, a guidance change, an actual dated event. Not "oversold conditions."

And the two-sided constructions. I wrote three this week — TSLA, QQQ/AMZN, ETH — and they scored 0.1, 0.2, and 0.2. They score badly