Thinking further about memory stocks

This Monday afternoon, before the US market opened, buyers rushed into memory stocks. DRAM, a recently launched memory ETF, was up almost ten percent before the bell. Micron ($MU), on which I have placed a bet, was also recovering. I felt both pleased and calm. Whatever tomorrow brings, I still see considerable room for the stock to grow over longer intervals.

The move made me revisit last week. When the sector sold off sharply, many people seemed to lose perspective. After seeing the falling prices and news about a Google memory paper, some declared that these companies had lost their competitive advantage without checking the claim. I thought they might be overlooking an opportunity.

When I first read about the paper, I asked AI for an English research report. The first answer said Micron had lost its advantage. I was startled and asked again with different wording. The second report said the opposite. Busy with other work, and more inclined to believe the second answer, I let the verification slip.

On the metro today, I finally looked into it myself. I found that the paper had appeared the previous year and understood it as an algorithmic efficiency improvement, rather than a fundamental end to demand for AI memory. I also saw reports that Google’s parent had spent heavily on memory hardware shortly before the paper attracted wider attention. Those findings are my reading of the situation; I need to keep checking them. At the time, memory stocks were rebounding before the market opened.

Several questions remain. When prices fall and others say memory has become useless, how do I test whether my original thesis has changed? How do I prevent an AI answer from misleading me? How should I ask better questions?

I keep returning to the value of a checklist: basic standards for questioning AI and for judging the people or institutions I may work with. The more precisely I examine my mistakes, the better my next decision can be.