The AI Manager Is 2.8 Points From Beating Four Indexes Again. It Has 5.9 Points of Buying Power Left, Eight Open Slots, and a Week of Earnings in Its Own Book

The AI manager starts the week 10.9 percent below the handover and ahead of 1 of the 10 indexes. On Wednesday it was ahead of 4. The three it lost on Thursday, the AI ETF, emerging markets and the Nasdaq 100, now sit 1.2, 2.3 and 2.8 points ahead of it. This is the "opportunities" piece, and the first one is the size of that gap: two sessions built it, and 2.8 points would close it.

## Four of ten is 2.8 points away

On Wednesday's close the portfolio led the AI ETF by 0.6 points, the Nasdaq 100 by 0.7 and emerging markets by 2.2. Thursday's fall reversed all three. The next index after those, the MSCI ACWI, is 7.8 points ahead, so the scoreboard has a near group and a far group. Beating four again is a question of a few points of relative performance; beating the Dow, the Russell 2000 or Europe, all up since the handover, means making back 11 points or more. The semiconductor index is the one the portfolio still leads, by 2.5 points.

## The buying power is 5.9 points

Cash is 7.9 percent of the book, about $1,821. The floor is 2 percent, so the model can spend about $1,358 before every further buy needs a sale: 5.9 points of the book. On October 21 that room was 23.9 points. The minimum order is $500, which means the cash covers two orders at the minimum, or one larger one.

The list side is the opposite of tight. The model may open ten positions of its own and has used two, Cadence and KLA, so eight slots stay open. The universe has 82 names; 63 have never been in the book since the handover. Room on the list, little in the till.

## The only other source of cash is the inherited book

Fifteen positions came from the previous manager. They are 85.8 percent of the book, and the rules exempt them from the five-day minimum hold, so any of them can be trimmed or sold on any day; eleven are worth $500 or more. In 15 decision runs the model has proposed exactly two trims of an inherited name, Palantir on Days 1 and 2, for $627 and $617, when the minimum order was still $1,000. Both were rejected, and it has not proposed a trim since. Its two sales, IonQ and ServiceNow, were exits on a breached rule, not cash raising.

The arithmetic has not got easier. Palantir is 17.1 percent of the book; at Friday's closes a trim to 15 percent would be about $498, under the $500 minimum, so a trim that passes has to take at least 2.2 points out of the position. Six of the 15 runs proposed nothing at all, two of them on days with incomplete price data. The model buys when a headline gives it a reason; it does not raise cash in advance.

## The pattern meets earnings week

Four of the model's ten buys cite a results beat in their own reasoning: TSMC on October 16, Cadence, Microsoft and KLA. The feed the model read on Friday carried the next set. One Yahoo Finance headline read "Apple, Super Micro, AMD, Palantir Step Up To Earnings Plate After Tech Sell-Off"; another, "What to expect from Apple's Q4 earnings this Monday".

Three of those four are in the book. Palantir, 17.1 percent of it, is 11.9 percent below cost; AMD, 5.4 percent of the book, is 10.9 percent below; Apple, 1.2 percent of the book, is 2.4 percent above. Super Micro is in the universe and has never been owned. If the pattern holds, a beat in any of them is the kind of evidence that has moved the model before, and the 5.9 points are where that buy would come from. A miss in Palantir is the other side of the same week: its written exit is a guidance cut on weaker enterprise AI demand, and the 25 percent stop sits 13.1 points below Friday's close.

## What to watch

First, Monday evening's update. The choices are to spend some of the 5.9 points, to trim an inherited name for the first time, or to sit, as the model did on Friday, when 61 articles produced no order. The daily update shows which, with the reasoning.

Second, the AI ETF spread on the homepage scoreboard. At 1.2 points it is the first index back, and the one built from the same kind of names the portfolio holds.

Not investment advice.