The Morning Bell No. 2: The Print Only Crypto Heard
- Branden Bell
- 2 days ago
- 6 min read
At 8:30 this morning the government revised second quarter core inflation up to 3.6%. The S&P 500 answered by moving four hundredths of one percent. Bitcoin dropped through $78,000, gold gave back 1.4%, and Strategy and Circle each fell about four and a half. The stock market did not react to the inflation print. Everything that runs on cheap money did.
Second thing, and I would rather say it than have you notice it. This edition is late. Yesterday I promised the next one would publish before the opening bell. It did not. The pipeline broke on the art step and I fixed it live, which is what running a newsletter on a system you are also still building looks like. Every number below is marked at 11:10 AM ET, mid session, not premarket.
I'm not a financial advisor and nothing in The Morning Bell is financial advice. This is entertainment and education based on research my AI system produces and things I may or may not be doing personally. The account featured here is a small agentic test account, fully separate from my personal trading. Always do your own research.
The tape after the bell
Marked at 11:10 AM ET:
S&P 500 (SPY): 766.20, +0.04%. Flat is the whole story.
Nasdaq 100 (QQQ): 710.41, -0.04%. Also flat, with Nvidia reporting after the close.
Small caps and bonds: quiet. IWM -0.17%, TLT -0.26%, and my volatility proxy is lower on the day.
Bitcoin: $77,661, -1.8%. Back under the $78,000 zone it reclaimed Monday. Gold (GLD) gave back 1.4% alongside it.
The crypto equities took it worse. Strategy (MSTR) -4.5%, Circle (CRCL) -4.5%, Robinhood (HOOD) -2.6%.
My system's regime call is amber. The reasons: the indexes are flat, volatility is falling and bonds are quiet, so this is not a risk-off day in any broad sense. The damage is concentrated in the assets that price off cheap money, which is exactly what a hotter inflation revision should hit first. It is not green either, because Nvidia after the close is a binary and Friday morning is the new Fed chair's first keynote.
One correction from yesterday, out loud. Edition No. 1 printed a 60% chance of a Fed hike this year, sourced from a newsletter quoting Kalshi. My own ledger flagged that number this morning as stale. CME FedWatch has September at 38% and the prediction markets are between 26 and 32%. The rule I keep re-learning is that an aggregator quoting a number is not the number.
What the system flagged overnight
Durable goods beat and nobody cared. Orders rose 1.1% against a 0.4% forecast, with computers and electronics up 3.1%. A genuinely strong manufacturing number that got buried, because it shared an 8:30 release with a hotter inflation revision.
My beats-get-sold read was wrong by lunch. The premarket note said retail earnings were beating and selling off anyway. That held for about an hour. Kohl’s is now +2.2%, Bath and Body Works +5.1%, and Abercrombie is up 37% on the day. Williams-Sonoma is roughly flat. Premarket is not the tape, and I published that read before it was tested.
Saylor changed the Strategy story. He said bitcoin sales are necessary for the credit business and that credit outranks buybacks, reported by Cointelegraph on Tuesday. My ledger scores that as thesis-changing rather than noise.
Three independent bears now point the same way on MSTR. Management's own words, the analyst my system weights highest saying the flywheel is dead until STRC gets back to par, and MSCI's open consultation on non-operating companies with an estimated $2.8 billion of passive selling attached. Different reasons, same direction. That is the best supported idea on my board today.
The setups on my screen
Research notes, not recommendations. Graded here tomorrow.
Momentum long, Tesla. $344 now. The entry band my system wrote is $344 to $352 with invalidation at $334. The Cybercab event in Austin on September 3 is a dated catalyst eight days out.
Research short, Strategy (MSTR). $121.16, down 4.5% today and already through the $126 to $130 fade band. See above for why three separate sources point the same way.
Options study, Nvidia. The August 28 215/225 call spread from yesterday marks at $2.72 against the $2.26 I published. Tonight settles it either way.
Pullback watch, Astera Labs. $284.61. The accumulate zone is $268 to $278 and it still has not come to me. A patience trade is only a trade if you actually wait.
The account behind all of this holds $115.32 in cash. None of those four are fundable at that size and I am not going to pretend otherwise. The point of publishing them is the record, not the fill.
Today's calendar, all times ET
8:30, done. Q2 GDP second estimate 1.5%, unrevised. PCE prices +5.3% and core PCE +3.6%, both revised up two tenths. Durable goods +1.1%.
13:00. Five year note auction.
After the close. Nvidia, plus Salesforce, CrowdStrike, Synopsys, Okta, HP and Veeva. Nvidia is the one that moves everything else.
Thursday. Jobless claims at 8:30, Jackson Hole day one, then Marvell and Workday after the close.
Friday 10:00. The new Fed chair's debut keynote. After tonight, that is the event of the week.
How I taught the system to grade itself
The report card at the bottom of this newsletter is a Claude skill, and it is the smallest useful thing I have built in a while. Here is the whole shape of it.
A skill is a folder with a markdown file in it. The file has a name, a description and instructions. That is the minimum. Anything else in the folder, scripts included, is optional.
The description is the hard part and almost nobody treats it that way. It is not documentation, it is the trigger. Claude reads descriptions to decide which skill applies to what you just asked for, so a vague one means the skill never fires and a greedy one means it fires when you did not want it. Mine names the situations out loud: grading yesterday’s setups, marking the report card, the daily newsletter run.
The instructions are the boring part, and boring is correct. Mine says: read yesterday’s published post to find the setups, pull the current mark for each one from the broker data, compare it to the entry reference that was actually printed, and write the row. Then it hands the rows to the engine that paints the board.
One rule I now put in every skill I write: a written definition of done and a written definition of give up. Without the second one it will keep trying, quietly, forever. That lesson came from a scheduled task that woke up thinking it was the wrong day, which I wrote up in more detail in how scheduled tasks work while your computer is asleep.
If you want to see a finished one instead of a description of one, the brand kit skill is free and ungated, no email. Same shape as this one, aimed at a different job.
The report card
Four setups from edition No. 1, marked from Tuesday's close:
Research short, Strategy. $126.83 to $121.16. That is 4.5% in favor of the short. A.
Options study, Nvidia. Published at $2.26 on the mark, now $2.72. Up 20% and still open. Tonight grades it properly. B.
Momentum long, AMD. $479.18 to $483.13, +0.8%. Right direction, small. B.
Pullback watch, Astera Labs. +0.7%, and it never traded into the zone, so there was nothing to do. No entry. C.
Every mark is green and the account traded none of them. That is two days running. Yesterday I wrote that publishing the record would be the fix for that. It has not been the fix yet, and I am not going to keep writing that line as though saying it counts.
One more thing I do not like. The account is $115.32, down $1.92 from Tuesday, with no open positions and no order I can find. It is small money and it is still an unexplained line, so I am chasing it before I add anything.
Where this leads
Building agentic systems like this one is what I do at Bell Consulting Solutions. Mostly for client delivery operations rather than trading, but the architecture does not change: connected data, skills that encode the judgment, a schedule, and a paper trail you can audit afterward. If you want to talk about one for your own business, reach out directly.
And the question for today. My system caught two of my own errors before lunch, the stale hike odds and a premarket read that did not survive to 11 AM. What is the mistake you would want caught and named for you every morning, before you had to notice it yourself?
One more time for the people in the back: not financial advice, entertainment and education only, based on what I may or may not be doing personally. See you tomorrow, and I am aiming for before the bell.
Sources:
Comments