Three questions stand out. Each fits the data's grain (one row per stock per trading day, 2012-01-02 to 2025-08-04) and the decisions the dataset seems built to serve. **1. Did the AI surge change price levels, or only a few leaders' price levels?** - Compare each company's trajectory before and after the ChatGPT launch (late 2022) and rank companies by how much they re-rated. - This shows whether the AI boom lifted the whole semiconductor sector or concentrated gains in a handful of names. That matters for deciding which stocks to overweight or underweight, and which to include in a sector basket. - There is enough data for it. The profiling queries found 152 companies and 422,729 rows, and 99,984 rows fall in the post-ChatGPT window, about 24% of the history. The median close across all rows is about $35.57. **2. Has risk changed? Is the intraday range wider in the AI era, and for whom?** - Compute the daily range as (high − low) relative to price, and compare it before and after the surge, by company. - The all-history average range is about 3.7% of price. The open question is whether that average hides a regime shift, with some companies far more volatile now and others calmer. - This bears on position sizing. It would also show whether the AI names carry a different risk profile from the legacy or cyclical ones. **3. When does unusual trading volume predict something, and did the surge change what "normal" volume is?** - Average daily volume is about 8.5 million shares. Flag days where a stock's volume is far above its own rolling norm, then look at what happens to price over the following days. - This tests whether volume spikes work as entry or exit signals. It also tests whether the post-2022 baseline shifted enough to make older thresholds misleading. - The data has a caveat. 4,888 rows have zero volume, which likely reflects halted trading, data gaps or illiquid tickers. Exclude or separately examine these rows before building volume signals. **Caveats before starting** - My profiling found 152 distinct companies and tickers, slightly more than the roughly 149 on the dataset card. Check for duplicate or renamed entities before ranking companies. - Companies have different start dates. Newer listings would need handling so they do not bias the before/after comparisons. I have not yet run these analyses. The figures above come only from the profiling queries, so the next step for any of the three is a company-level before/after query.