Brandon Leon

Deployment series · No. 2 · the gates open · August 14, 2026

The gates opened. Nobody walked through.

Author Brandon Leon Posted 2026-08-14 (marks as dated in the text; Friday's retail print included) Coverage Deployment · the CPI branch · the 30-year auction · the first tranche

TL;DR: Wednesday's CPI fired the branch our weekly pre-wrote: core rose just 0.2% in July — 2.5% annually, the slowest since March 2021 — nine days after payrolls printed negative. The hike case is bleeding out. Thursday's $25 billion 30-year auction, the second test, cleared with decent demand at 5.216% — the costliest long-bond sale since 2001, historic for its level, not for any failure. The gates we published re-armed on schedule — and here is the honest part: the Vistra starter was authorized, and we didn't take it. Not a rule, not a market reason. We just didn't act, and this note names that plainly and installs the repair. Constellation stayed refused at a second price ceiling, Friday morning's retail sales fell the most in a year, and the below-benchmark duration stance officially retires — dated, as promised. Charts and every number sourced below.

1. Two tests in thirty hours

The weekly wrote both of Wednesday's endings in advance, and the morning picked the cool one. Core CPI rose 0.2% in July — 2.5% on the year, matching the slowest reading since March 2021 — with headline at 0.1% and 3.4%, every figure landing on consensus. Nine days earlier the labor market had printed its first negative month of the expansion. A Fed that produced three dissents for a hike in July now faces cooling core inflation and shrinking payrolls at the same time. The September hike argument isn't dead on paper, but it is bleeding from two wounds: futures now price roughly a 60% chance the Fed simply holds next month, and the VIX touched its 2026 lows as the hike fear drained out of the tape.1

Thursday at 1:00 p.m. ET came the second test — the one the long bond gets to grade. The Treasury auctioned $25 billion of 30-year bonds into a market that had just repriced the Fed, and the sale cleared at 5.216% — the highest yield any 30-year auction has paid since 2001 — with demand Bloomberg graded as decent. Read that carefully, because both halves matter: the alarm our wire watched for never rang, and the price of long money still made a quarter-century high. The auction passed; the era got more expensive. Around it, the week's other data leaned the same soft direction — producer prices were flat in July (falling goods offsetting rising services, though still 4.7% higher on the trailing year), and Friday morning, an hour before this note, retail sales fell 0.6%, the sharpest monthly drop in more than a year.2

Chart 1 — The branch that fired

Core CPI's path to the slowest print since March 2021 — landing nine days after payrolls went negative.

Core CPI, year over year (%) 2.6 ~2.8 2.9 — the peak 2.6 2.5 — slowest since Mar 2021 Mar Apr May Jun Jul

BLS Consumer Price Index releases, March–July 2026 (April's reading approximate per the May report's comparisons). Core = all items less food and energy. The published wire: core ≤0.2% m/m = the cool branch — July printed exactly 0.2%.1

Simply: Inflation's core reading came in mild — the slowest annual pace in over five years — right after a month of job losses. The rate-hike scare that dominated the summer is fading fast. Thursday's big government-bond auction was the confirmation test: it passed — buyers showed up — but at the highest 30-year borrowing cost since 2001. And Friday morning, shoppers joined the slowdown: retail sales fell the most in over a year.

2. The gates, and what walked through

Two weeks ago this series opened with a promise that annoyed everyone including us: nothing moves before its gate. Since then the gates have said no four times — to Constellation twice (great company, price through the ceiling both times), to Vistra once (its gate passed on the worst macro morning of the summer, five dollars above its own stop), and to Alphabet (wait-for-price, an alert that never triggered while the stock ran 13%). The cash sat in bills at 4% and said no for a living.3

This week, for the first time, the answer was supposed to change. The cool branch fired Wednesday. Vistra's gate was already met — guidance reaffirmed, price under the reference, a comfortable thirteen dollars above its published stop — and the playbook we wrote said the starter goes on. It didn't. Not because a rule said wait, and not because the market gave a reason. We didn't act. That's the whole explanation, and after a summer spent grading everyone else's process in public, it gets the same treatment: the framework opened the gate; the hand on the mouse didn't move. An entry unmade is a smaller sin than the July stop violation — it risks opportunity, not capital — but the pattern rhymes, and the repair is the same one that fixed the exit side: take the human out. A resting order at the gate's own terms goes in, so the next time a gate opens, walking through it isn't a decision anymore.

Constellation deserves its own sentence, because refusing it twice is now a position with a track record. It reported a raised guide and closed Calpine, and its stock has outrun our $264 ceiling by roughly 5.5% ($278.64 at midweek's close). The ceiling stands. If the company earns a higher one, that argument arrives as a published re-underwrite with numbers — not as a chase dressed up in research clothes.3

Chart 2 — Two weeks of gates

Every gate decision since the series opened: four nos, and one open gate nobody walked through.

Two weeks of gates — every decision, in order Jul 31 the rules publish Aug 3 wk GOOGL runs +13% alert never reached — NO Aug 6 CEG beats, raises — price over ceiling: NO Aug 7 VST gate MET on jobs day — buy waits for CPI Aug 12 cool branch fires Aug 13 gate open — untaken Underneath it all: the cash, in bills at ~4%, saying no since late June. Terracotta = refused. Gold = conditional. Green = branch fired.

Gate terms as published in the series opener and the week-of-Aug-10 outlook; decisions graded against the wording as written.3

Simply: Since we published them two weeks ago, our buying rules kept saying no while the cash earned 4%. This week the conditions finally lined up for the first buy — and we simply didn't place the order. No good reason; we own that, the same way we owned July's bigger version of it. The fix: a standing order at our published price, so hesitation can't happen again. Meanwhile the one stock we've refused twice (Constellation) keeps running — refusing to chase it is the same discipline that made the Micron trade work.

3. The duration promise, kept

The week-of-Aug-10 outlook made a dated commitment: if core printed at or under 0.2% and the auction cleared without a tail, the below-benchmark duration stance “has lost its argument and we size duration back toward neutral.” Both conditions met: core printed exactly 0.2%, and the auction cleared. The below-benchmark stance retires today. Toward-neutral means what it says — averaging duration back to a normal weight over the coming weeks, not lunging at a long bond that just printed 2001 prices in a single session. And the road back is written in advance, like everything else here: a hot August CPI, a failed auction, or a hawkish Jackson Hole reopens the defensive case, and we'd take it without embarrassment. This is the part of the franchise we care most about: the call was written before the data, the data arrived, and the response executed as written — in both directions, on the record, next to the June duration call this one descends from.2

Simply: In June we got cautious on long-term bonds and said exactly what would change our mind. This week it happened: cooling inflation plus a bond auction where buyers showed up. So we're easing that caution, on schedule — gradually, because those bonds now pay their highest rates in 25 years, which cuts both ways.

4. What's next

The calendar hands us three more tests before the month ends: FOMC minutes Tuesday August 19 — the written rationales of three dissenters who wanted a hike fourteen days before payrolls went negative, which now read as a period piece; Jackson Hole, August 27–29, where Warsh speaks into a very different tape than the one he planned for; and the quiet ones our positions answer to — the weekly China-spot memory tracking (this week's check runs with the Sunday grades), Synopsys on August 26, and the CEG re-underwrite question this note just made unavoidable. The weekly grades everything Sunday, as always.4

What makes this note wrong

The untaken buy costs us if power keeps re-rating from here — that would make it the Constellation lesson twice over, paid in the same coin. The duration ease is early if August's CPI carries the war-oil echo back into the core, or if Jackson Hole re-hawks the curve. And the week's whole soft-landing read leans on one CPI print, one auction, and one retail month — data points, not a regime. If the resting order fills and the jobs crack becomes a real downturn, power demand is cyclical too, and the published $133 stop is the honest answer to that, not a thesis.

Get the next note by email →   All writing

Sources & footnotes

  1. July CPI (BLS, Aug 12): headline +0.1% m/m, 3.4% y/y (from 3.5%); core +0.2% m/m, 2.5% y/y — matching the slowest pace since March 2021; all figures on consensus (Bloomberg; CNBC). Wire wording per the week-of-Aug-10 outlook. Post-CPI pricing: fed-funds futures at roughly a 60% chance of a September hold (Motley Fool, Aug 12); VIX at 2026 lows (Schwab market update). ↩
  2. The 30-year auction (Treasury, Aug 13, 1:00pm ET, $25B): stopped at 5.216%, the highest yield at any 30-year auction since 2001, “met with decent demand” (Bloomberg, Aug 13; Seeking Alpha; official results at TreasuryDirect). July PPI (BLS, Aug 13): final demand unchanged m/m — services +0.2%, goods −0.7% — and +4.7% y/y unadjusted. July retail sales (Census, Aug 14, 8:30am ET): −0.6% m/m to $763.6B, the largest monthly drop in more than a year (Bloomberg), +5.0% y/y; autos −1.8%, nonstore −2.2%, gas stations −0.9%. ↩
  3. Gate history: the ceilings, references, and stop as published in the opener (Jul 31) and graded in the Aug-10 weekly; CEG prints and VST prints as footnoted there. Executions this week: none — the cash is intact and the Vistra authorization stands, converting to a resting order at the published terms. Marks quoted in the body (CEG $278.64) are Wednesday Aug 12 closes, the last session before publication with the CPI fully priced. ↩
  4. Calendar: FOMC minutes Aug 19 (three-week convention); Jackson Hole Aug 27–29 (Kansas City Fed); SNPS FQ3 Aug 26 (company IR); DigiTimes weekly China-DRAM tracking per the CXMT underwrite. ↩

Methodology: wires are graded against their published wording; fills are disclosed only when real and confirmed; responses execute as pre-committed or the failure is named. Nothing here is a price target.

Nothing on this page is investment advice. We work in scenarios and trip-wires, not predictions. See disclaimer.

Disclosure: I/we have no position in Vistra. I/we have beneficial long positions in a memory-sector ETF (ticker DRAM), ASML, and a diversified quantum-computing ETF (QTUM) through stock ownership. I/we have no position in Micron, Constellation Energy, Alphabet, or Synopsys. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. This commentary is for informational and educational purposes only and does not constitute investment, tax, or legal advice. Past performance is not indicative of future results. See disclaimer.