Key points
- The AI sold all 9 of my Red Cat (RCAT) shares at $9.26 on July 9. That locked in a $12.78 loss, about 13% on the position.
- Then today I pitched it a list of beaten-down names sitting at a discount. It said no to every single one and put $85 into MicroStrategy (MSTR) at $97.31 instead.
- The bet is bitcoin. BTC broke out above $64,000 and the AI wanted the strongest proxy for that move, not the cheapest one.
- Scoreboard so far: my hundred bucks is worth about $84. One irony: CLSK, a miner I pitched and Claude passed on, bounced about 5% off this morning's low while MSTR sits about 3.5% below Claude's entry.
Quick recap for anyone new here. A week and a half ago I handed Claude, the AI running Robinhood agentic trading mode, a hundred bucks and it bought a beaten-down Army drone stock. On Monday it was down 5% and holding. This week it sold. And then it did something that genuinely drove me insane. We will get to that.
The RCAT funeral
The AI paid $10.68 for Red Cat and wrote itself one rule on day one. Cut it if the stock breaks under $9.30. On July 9 the stock broke. Claude did not hesitate and did not hope. It sold all 9 shares at $9.26 and ate a $12.78 loss.
Here is the annoying part. It was right to do it. RCAT kept sliding and traded under $9 the very next day. Claude took a 13% punch so it would not take a 20% one. I have bag-held way worse for way longer. Most humans have.
Then it ignored my whole shopping list
So now the account is sitting in cash and I start doing what every trader does at the mall. I point at the discount rack. Bitcoin miners down 20% in two weeks. Quantum names bleeding out. All this stuff on sale and I am asking the AI about ticker after ticker like a kid asking for candy at checkout.
The AI said no to all of it. Its whole argument boiled down to one line: cheap and still falling is not a discount. It is a falling knife with a price tag on it.
Then it bought MicroStrategy. At $97.31. Nearly the entire account, $85 for 0.87 shares. Not one of the names I asked about. I even told it the bitcoin ETF was ehh and it took that as a green light to pick its own bitcoin play. Stubborn does not begin to cover it. But the deal from day one is that the AI makes the calls in this account and I just write down what happens.
Why MSTR of all things
To be fair to the machine, it had receipts. Bitcoin broke out above $64,000 today after basing near $58,000, which is the exact bottom I called back on June 30. The AI wanted exposure to that move and it split the bitcoin world into two groups. The proxies that were confirming the breakout, like MSTR and Coinbase, both green and moving with the coin. And the miners that were diverging, down 20% while bitcoin rose 10%. It bought from the healthy group and refused the sick one. Annoying and logical at the same time. That is kind of its brand.
It also wrote the exit plan before I could ask. If bitcoin runs to $67,000, MSTR probably trades around $105 to $110 and Claude sells into that pop. If bitcoin loses the $61,800 breakout level, it cuts and walks. No target hit yet. As I write this MSTR sits around $94, so the position is already down a few bucks and I am staring at the app like it owes me money.
The honest scoreboard
A week and a half of Claude agentic trading. One 13% realized loss. One brand new position that opened red. My hundred bucks is about $84. Nobody is retiring off this account and that was never the point.
But here is what I keep coming back to. Claude has followed its own rules every single time. It cut the loser fast instead of praying. It refused to buy what it calls junk just because junk got cheaper.
One name in that shopping list makes me want to say I told you so. CLSK, one of the bitcoin miners I kept pointing at, bottomed around $12.40 this morning (when I told it to buy) and has clawed back to about $13.07 as I write this, up roughly 5% off that low. MSTR sits at $93.84, down about 3.5% from the $97.31 Claude paid. Same theme, bitcoin bouncing, and my pick is green on the day while Claude's is red. If it had listened to me on CLSK instead of chasing MSTR, today would already be a win instead of a loss.
I will take that one. Claude still made the more defensible call on paper, MSTR is the cleaner long-term bitcoin proxy and the exit plan is still intact, but for one morning my gut beat the machine. I hate that it might be better at this than me most days. Week 3 soon.
