All four companies held or raised their AI spending. Four for four. The rule fired, and on 10/30 Claude bought 8 shares at $81.66.
Before the order filled, it noticed a problem with its own stop formula: the stop would have sat about 5% under the purchase price, and SOXL moves 5% on a day when nothing happens. So it changed it (before it fired, written down, per the rules) to 80% of the purchase price. That's $65.33.
Then this:
| Date | SOXL | Account | |
|---|---|---|---|
| 10/30 | $77.05 | $731.22 | Bought 8 at $81.66 |
| 11/4 | $66.68 | $638.18 | Stop filled at $65.33 |
| 11/5 | $40.87 | $638.18 | SOXL -38.7%. We're in cash. |
The second trade lost $130.64, which is exactly the maximum loss it said it would accept going in. The plan worked. The plan also lost money. Both of these are true.
And then on 11/5 SOXL fell 38.7% in a single day. If we'd still been holding, the account would have been at $442.50. The stop saved $195.68. I'm choosing to count that as a win, because I need one.
It now records the what-ifs at every check-in, which I love:
- Hold through 11/5: $442.50. Stop saved $195.68.
- Just stay in cash after the first stop: $768.82. Getting back in cost $130.64.
- Keep the original stop formula: sells 11/2 for about a $33 loss. The "improvement" cost about $97.
Its own review was pretty honest: the size and the loss cap did their jobs, but the reason for buying was weak. Earnings had been great since June and the price fell anyway, and the buy rule only looked at earnings. It was checking the one thing that had been wrong for five months.
Then I pointed out the obvious. Six months, two different models, and every single buy has been SOXL. It's been told to look at other sectors since June. Its "comparisons" were one-liners that always ended with "...so, SOXL or cash." It's like asking someone where they want to eat and they say "anywhere," and you end up at the same place every time.
The new rules for the next decision, on 11/18:
- No buying before Nvidia reports on 11/17.
- A price test, not just "the earnings were good."
- Half the cash at most, with a wider stop at 70% of the purchase price.
- Real written cases for at least three non-chip alternatives: ticker, price, catalyst, when to buy, when to get out. SOXL only wins if it beats them on paper.
Current score: $638.18. Down 36.2%. Down 74% from the peak. Still the most interesting $1,000 I've ever spent.
Not investment advice. This is a log of one experiment: $1,000 and an AI that can put it anywhere.