Key points
- Claude AI is 0 for 4. My $100 is now $74.59, down 25.4%.
- Two more stop-outs last week: Rigetti (RGTI) and D-Wave (QBTS), $7.06 combined.
- It passed on IREN at +17% over a failed July 6 pop, then watched IREN run to $40.13.
Four weeks in and my $100 experiment finally has a record. It is 0 for 4.
Here is the damage. Red Cat (RCAT) lost $12.78. MicroStrategy (MSTR) lost $5.57 and we wrote about that one because it gapped up 6% the morning after Claude AI sold it. Last week added two more: Rigetti (RGTI) for $4.90 and D-Wave (QBTS) for $2.16. My hundred bucks is now $74.59, down 25.4%, and I am not dressing that up.
Something changed over the last four sessions though.
The D-Wave trade lasted two hours
Claude AI bought 4 shares of D-Wave at $17.74 Thursday morning and it had a real reason. The stock double bottomed at $17.45 and reclaimed that level on rising volume, so the stop went at $17.20 just under the low. Two hours later it filled.
Here is the part I keep chewing on. The setup got followed correctly and it lost anyway. A 3% stop on a stock that swings 5% on a boring day is not risk management, it is a coin flip on noise. Tighten the stop far enough and you stop turning one big loss into no loss. You just turn it into a bunch of small ones.
Then it stopped trading
Friday the chip trade came apart. The VanEck Semiconductor ETF (SMH) fell 4% and bottomed at $536.81 on 22 million shares, roughly double a normal day. Claude AI scanned about 30 names and bought nothing. There was not one green name on the list.
The IREN trade it almost made
IREN (IREN) filed an 8-K Monday morning raising its year-end AI cloud revenue target from $3.7 billion to over $4 billion on $2.8 billion of new contracts with Microsoft, Nvidia, Perplexity and Figure AI. The stock jumped 17% and dragged the whole miner-turned-AI-landlord group with it. Hut 8 (HUT), Cipher (CIFR) and CleanSpark (CLSK) all jumped double digits.
Bitcoin did nothing, by the way. The iShares Bitcoin Trust (IBIT) was flat. This was an AI datacenter move wearing a bitcoin miner costume.
Claude AI had the order built. One share at $39.45.
I asked why it was buying the high of the day when its own rule says buy dips, and mentioned that IREN has a dilution problem. I also told it I was not going to stop it.
It went and looked instead. On July 6, IREN went up 13% to close at $43.91 on nearly identical AI infrastructure news. Then it gave the whole move back over the next nine sessions, closing at $33.62 on Friday. Anyone who bought that news day was down 23%.
Same stock. Same story. Two weeks earlier. It did not work.
The financials did not help either. IREN's quarterly revenue has fallen from $240.3 million to $184.7 million to $144.8 million while net income swung from a $384.6 million profit to a $247.8 million loss. It closed $3 billion of convertible notes in May. At a $14 billion market cap, that is roughly 24 times a shrinking revenue line.
Claude AI passed.
How it looks so far
Honest scoreboard, and it flatters nobody. IREN faded to $38.26 by 11 am, which made the pass look sharp for about an hour. Then it turned around and printed $40.13 just before 1 pm.
That is above the $39.45 where the order would have filled, so as I write this, passing has cost me money rather than saved me any. It also means IREN tagged the exact $40 level Claude AI named as its buy trigger, a couple of hours after naming it.
What I actually think
I am not thrilled about paying tuition four times running. But this is the first time Claude AI talked itself out of a trade using evidence instead of vibes, and it did it on the best-looking chart on my screen. That beats the failure mode that cost me $25.
It wants IREN to hold above $36.40 and clear $40 on volume before it will touch the thing, and the stock is testing that as I type. So next week is simple. Either it follows its own rule, or it finds a fresh reason to sit on its hands. One of those is discipline, and the other is just being scared, and I do not know which I am getting.
We are also running five AIs against each other on a pretend $10,000 if you want more of this. Check back next week. We will know either way.
This is general market commentary and opinion, not investment advice. Markets can go down as well as up, and you can lose money. Always do your own research and consider speaking with a licensed financial professional before making any investment decision.
