The three most useful questions combine the dataset's long history (2012-01-02 to 2025-08-04), its breadth (152 companies and tickers), and its daily price and volume fields. The profiling queries I ran show the figures below. **What the profile shows** - There are 422,729 stock-day rows. - The median close is about $35.57. - Average daily volume is about 8.47M, but median volume is only 898,500, so a few very heavily traded names skew the average. The maximum single-day volume is about 3.69B. - The average daily high-low range is about 3.69% of price. - Rows since ChatGPT's launch (Nov 2022) total 99,984, or 23.7% of the data. Rows from 2023 onward total 96,711. - 4,888 rows have zero volume. **1. Did the AI surge change how semiconductor stocks behave, and for whom?** - *Question:* How do price level, daily volatility (high-low range as a share of price) and volume compare before and after late 2022, company by company? - *Why it's interesting:* About 24% of the rows fall after the AI boom. That is enough history to test whether the gains were broad across the sector or concentrated in a few names. It would also show whether volatility rose along with prices. - *Decision it serves:* whether to overweight or underweight a name, or enter or exit a position. **2. Which companies carry the most risk for the return, and has that ranking shifted?** - *Question:* Ranking companies by return since 2012 against volatility (average range %, or the standard deviation of daily returns). Which ones were consistently efficient, and which only looked good during the recent run? - *Why it's interesting:* With a 3.69% average daily range, risk clearly varies across 152 names. Comparing the pre-AI and post-AI rankings shows whether leadership is durable or was a one-time re-rating. - *Decision it serves:* position sizing and the choice of names for a sector basket or benchmark. **3. When does unusual trading volume signal something, and is it tradable?** - *Question:* Can volume spikes (for example, volume well above a stock's own trailing median) be used to predict the next day's or week's price move or volatility? And do these signals behave differently in the AI era? - *Why it's interesting:* The gap between median volume (898.5K) and mean volume (8.47M) shows that volume is very heavy-tailed. Comparing each stock with its own baseline, rather than with absolute volume, could separate real interest from noise. - *Decision it serves:* timing entries and exits. - *Data caveat:* the 4,888 zero-volume rows look like halted days, illiquid names or missing data. They should be excluded, or at least checked, before any volume-based analysis. If you want, I can start on any of these. The pre-aggregated gold tables by company and year should make question 1 quick to answer.