The AI Manager Lost 10.3 Percent in Three Days. What Still Works: a Second Stop-Loss the Model Endorsed, One Buy on Evidence, and a Lead Over the Chip Index That Survived Three Red Days

The AI manager fell 2.1 percent on Monday to $22,675, its third red day in a row, and the three together cost 10.3 percent. Since the handover on October 9 it is down 12.7 percent and ahead of 1 of the 10 indexes. This is the "what is working" piece on the worst stretch of the AI era: the list is short, and each item sits next to the loss it did not prevent.

## The stop-loss fired a second time, and the model agreed

Datadog closed Monday at $216.10 against an average cost of $293.26, 26.3 percent below it. The 25 percent stop queued a sale for Tuesday's open, about $304, the whole 1.3 percent position. The new part is the model's own verdict. Its exit check still called the Datadog thesis intact (one customer's usage cuts being too little evidence), then it proposed the same sale itself under the price rule; the rules engine set that aside only because the forced order was already queued. ServiceNow on Friday went the same way. Two stops in three sessions is the rule doing what it says, and two inherited positions sold a quarter below cost: $457.28 realised on ServiceNow, and Datadog leaving ahead of the results report the feed previewed on Monday under "Datadog Before Q3 Earnings: What's in the Cards for the Stock?".

## One buy on evidence, with the cash kept

With 63 articles in front of it the model made one choice of its own: add to ASML, from 5.66 percent of the book to 8.2, about $576. The reason is a management line the feed carried on Monday, "ASML Fell 6% in Intel’s Chip Rout. Its CEO Says Supply “Will Not Fully Meet the Demand”": machine supply will not fully meet demand in 2027 and 2028. The exit is written: out if expansion plans or guidance are cut because chip-equipment demand deteriorates. The previous ASML add, October 23 at $1,550.48, is 1.6 percent under its fill, the smallest deficit among the eight buys under water. Cash stays at 8.0 percent, about $1,821, before Tuesday's orders. The habit since October 16 holds: the model buys on a reported fact and names the fact.

## The chip spread shrank, then rebuilt

The portfolio is ahead of one index since the handover: SMH, the semiconductor index, down 15.8 percent against the portfolio's 12.7, a spread of 3.0 points. The three red days moved it both ways. On Thursday the book fell further than the chips and the spread shrank from 3.2 to 0.4; on Friday and Monday the chips fell further and it rebuilt to 2.5, then 3.0. Over the three days the book lost 10.3 percent and SMH 10.5, so a book with 41.4 percent in chip names has, on balance, fallen a little less than the index built from them. It is a thin lead and the only one on the board: the AI ETF is 1.5 points ahead, the Nasdaq 100 3.4, emerging markets 4.2.

## Cadence is still the one buy above its fill

Of the model's nine live buys, one is worth more than it cost: Cadence, bought October 27 at $325.96 and closed Monday at $329.51, up 1.1 percent, 3.4 percent of the book. The other eight are below their fills: ASML 1.6 percent, KLA 3.9, the two TSMC adds 5.4 and 5.0, Microsoft 6.1, Shopify 10.4, Snowflake 10.5, AMD 10.9. Across the nine, $9,100.66 of buys at fill are $447.92 under water, and the two sales have realised $525.31 of losses. None has breached its written exit.

## What is not working

Three of 17 positions are above cost: Monday.com at 7.1 percent (0.1 percent of the book), Cadence 1.1, Apple 0.5. Alphabet is 20.1 percent below cost at 8.9 percent of the book, 4.9 points from the stop line. AMD is 17.6 percent below, Snowflake 16.4, Meta 15.6. Palantir, 17.0 percent of the book, closed Monday 14.7 percent below cost and reported after the bell under "Palantir Earnings Top Views. Revenue Accelerates For 13th Straight Quarter."; the share reaction prints on Tuesday.

## What to watch

First, Alphabet against the 25 percent line, 4.9 points away. The rule has fired twice; the model endorsed it both times.

Second, whether the chip spread survives a green day. It rebuilt because the book fell less than SMH on Friday and Monday; the test is a rebound, when 41.4 percent in chips has to rise as fast as the index. The homepage scoreboard shows the spread after each close.

Not investment advice.