My current view is that GPT 6 Astra is not meaningfully better than Fable 5.1 for my personal work, but that using both side-by-side is nonetheless very helpful and additive. I have been using GPT 6 Astra and Fable 5.1 a bunch over the past two days, largely for policy analysis, memo writing, and simpler software (e.g., making dashboards and forecasting models) that still nonetheless seems difficult conceptually. Across a variety of tasks I've done, it's been fairly random and hard to predict…