A lot of what we accept in life is determined by how expensive it is to investigate something better.We stay with a service we don’t especially like. We keep doing a piece of work in a way that annoys us. We wonder whether a different city would be better for our family, or whether an idea we’ve had for years could become a business. Some of those thoughts become decisions. Plenty of them never get that far. We can see the work we’d have to do just to find out, and there isn’t room for it this week.There wasn’t room last week either. Eventually, we stop bringing it up.Astra, OpenAI’s new GPT-6 model, gives us a reason to reconsider those choices. It is better at using computers and carrying an assignment through changing circumstances. The practical question is how much of the work between asking and receiving a useful result we can now leave to it.I firmly believe that maximizing choice leads to a richer life.Tim Urban’s life-paths diagram makes that belief easy to picture. Behind the point marked today, a single green line traces the life you have lived, surrounded by the paths you didn’t take. Ahead of today, the green line branches into many possible futures.Diagram by Tim Urban of Wait But Why.A possibility you cannot afford to investigate is barely a choice at all. It can exist in theory while remaining out of reach because finding out whether it would work takes more time, money, or help than you have. Making that investigation less expensive gives you a wider set of futures you can seriously consider. That is what interests me about Astra: more of those green branches becoming practical to explore.If you still have to organize every search, move the information between applications, and tell the model what to try next, the investigation remains your job. You have help with pieces of it. When an agent can carry more of that work, a question that once demanded your weekend may become worth asking.I want to consider what changes when we start handing Astra work we weren’t going to do at all. A faster model can save time on an existing task. A more capable agent can change the set of possibilities we live with. The things we never attempted don’t appear on a timesheet, but they can matter more than the things we finish faster.This piece covers:23 recipe cards for the work you keep postponing. The renewal you haven’t read, the subscriptions you’ve never audited, the quote you never got around to comparing. Each card is a working prompt you can paste as written.What the benchmarks actually license. Astra more than doubles its predecessor on business-workflow tasks, and 41.4% is still not dependable completion. Here is what those numbers do and don’t justify trying.The assignment to give before you’ve decided. “Help us move” assumes the decision is made; “help us understand whether moving would improve our lives” is different work, and it can succeed by telling you to stay.What your customers are actually paying for. Some of them are staying because leaving would be another job, and that reason is now open to automation in both directions.How a narrow search can look thorough. The agent decides which alternatives you ever see, so the useful instruction is to have it investigate what would make its own recommendation wrong.Below: what the evidence actually supports, and the six places I would point Astra first.