The AI Manager Made Its First Two Trades on Its Worst Day Yet. Here Is What It Still Has to Work With.
Day 5 of the index race: the AI-run paper portfolio fell 3.1 percent in a day, is down 5.3 percent since inception and beats 4 of 10 indexes. The same day it added to TSMC and ServiceNow, its first trades ever, after a fix that took its daily news input from one article to 62. Three things now work in its favour: a model that can finally see its universe, cash still at 26.7 percent of value, and a written exit condition on all 17 positions.
Four trading days of nothing, then two orders on the worst day so far. The AI manager added to TSMC and ServiceNow at the open on October 16, and by the close the portfolio was down 3.1 percent on the day. Both facts belong in the same sentence, because the race is scored on the close, not on the activity.
## The scoreboard on Day 5
As of October 16 the portfolio is worth $24,589, down 5.3 percent since inception. It beats 4 of 10 indexes: SMH (down 10.5 percent, a spread of 5.2 points in the portfolio's favour), AIQ (down 6.9 percent, spread 1.5), EEM (down 5.9 percent, spread 0.6) and QQQ (down 5.5 percent, spread 0.2). It trails IWM (down 3.0 percent, 2.3 points against), SPY and ACWI (each down 2.0 percent, 3.4 against), URTH (down 1.3 percent, 4.0 against), DIA (up 0.6 percent, 5.9 against) and VGK (up 1.5 percent, 6.8 against).
On Day 3 the count was 5 of 10. The small-cap index flipped to the other side. The lead over the chip index is still the largest spread on the panel, and it is still a lead the model did not earn: this book holds less of what fell.
## Opportunity one: the model can see its universe
On Day 3 the daily decision was made on one news article. On Day 4 the run received 61 and on Day 5 62. The difference is a feed of per-ticker headlines for the 82 names in the universe, which now supplies 60 of the items the model reads each day; the site's own news feed supplied one and two.
The first unforced trades followed within a day. The TSMC add, 3.9521 shares at $421.22 for $1,664.70, cites one headline: the third-quarter beat and raised guidance. The ServiceNow add, 7.7772 shares at $151.92 for $1,181.51, cites the EmployeeWorks launch, and the model's own reasoning limits the size because product launches "do not yet establish incremental revenue". Both were decided on October 15 and filled at the next open. After one close, both sit below their fill price.
## Opportunity two: the cash
Cash was 37.1 percent of value on Day 4 and 26.7 percent after the two fills. The human left that cash. The model has now used part of it, in two names it already held, on a day the sector sold off. Buying weakness with inherited cash is a decision; so is the quarter of the book still waiting.
## Opportunity three: every position has an exit written down
The automatic stop-loss, 25 percent against average cost under the v2 rules, is the only exit the code enforces. As of October 16 each of the 17 positions also carries a thesis and a written invalidation on its holding page, and the daily run checks them: the Day 5 update records two checks, both intact. TSMC's exit condition is a cut to current-quarter guidance; ServiceNow's is declining enterprise adoption. Those are the lines to hold the model to.
## What is not working, said loudly
Only two of the 17 positions are above cost: Apple, up 2.0 percent at 1.2 percent of value, and Monday.com, up 1.6 percent at 0.1 percent. IonQ is down 21.6 percent against cost, the closest to the automatic stop. Snowflake is down 16.8 percent and Datadog 15.7 percent. Palantir, the largest position, is below cost as well.
## What to watch
First, the two new exit conditions. If TSMC's guidance or ServiceNow's adoption numbers turn, the model has to call its own invalidation "triggered" in public, and the update page will say whether it queued an exit or declined to.
Second, IonQ against the 25 percent stop. At 21.6 percent below cost it is the position the code, not the model, is most likely to sell next. The trade log will show which of the two acts first.
Hypothetical paper portfolio. Not investment advice. No real money is traded; decisions are made by an automated AI system.