Every call, graded in public.

The call

Opened ~$668 (May 6). Rated BUY, then HOLD. Base target walked $1,500 → $1,300 → $1,100 across three public cuts, each narrated. Hard $900 close-stop.

The mark

$979 — ~+47% from entry, with ~60% of the position booked into a $950–$1,200 scale-out and the runner held above the $900 stop.

The position is the win; the calls inside it are the honesty. I said the FQ3 print was priced and would fade — it ripped ~16% to a record $1,213.56 instead, and the note opened with “I was wrong on the direction, full stop.” Then I cut the bull target three times as the SK hynix listing and added industry capacity broke Micron’s scarcity premium — a multiple cut, not an earnings cut, argued each time. Right on the position, repeatedly humbled on the calls.

The miss I owned → Latest re-underwrite →

The call

Initiated HOLD (Jul 5) — fair value ~$115, 12-month target $135. “Sell the premium, buy under $120.” No position.

The mark

~$145 — down from the ~$214 draft-date high we refused to chase, closing the distance to our $135 target.

We initiated the Street’s only rating below the stock — seven bank targets ($205–$300) sat above it. The melt-up has since round-tripped ~32% from ~$214 toward our fair value, exactly as the note argued it would. It isn’t closed, and the supply cliff — lock-ups releasing through December — is still ahead. But the contrarian call is working.

The initiation →

The call

Initiated HOLD (Jul 2) — probability-weighted reference ~$340, accumulate lower on a $320 / $285 / $250 ladder, off a $371 anchor. No position.

The mark

$435 — ~+17% above the anchor, above a ladder it never triggered, trading toward the $484 Street target we’d called the bull case.

Wrong so far, and I’ll say it as plainly as I’d say a win. The call was patience — wait for weakness to accumulate into. Instead Dell ran to the upside on exactly the AI-server durability the note argued the market hadn’t yet proven. The ladder still stands and the memory-cost kill-shot is still live, but the entry thesis hasn’t paid. This is the one in red.

The initiation →

The call

Pre-meeting: Warsh underwhelms the hawks, the surprise skews dovish-of-fear, the rally holds.

What happened

Wrong. He out-hawked the tape — the 2-year ripped 16bp to 4.216% (the exact 15bp-plus falsifier I’d named pre-call), the dots flipped to a 2026 hike, the S&P fell to 7,420.

Killed by my own rule, the same afternoon: “We were wrong… we own it.” The lesson I took from it — a new chair’s first meeting is structurally hawkish, and you do not fade a risk that’s cleanly aligned with the actor’s incentives — is the one the July 8 minutes went on to confirm. A wrong call that paid for itself in a framework.

The grade →

Three calls from the prior week, scored in print the week after, in my own words. The jobs base case — half credit: “the hike moved out as we said, but +57K is not a middle number, it’s a soft one.” “Suppliers reprice last” — the miss, owned in full: “on price, the split trade lost money in week one; that’s the grade, and there’s no dressing it up.” And the “calmer long end” — leaking, not yet broken: “the week traded Schwab’s scenario, not ours, and we say so.”

The scoreboard → How the rules held next →