A year with Codex, a week with Astra

I have been using Codex actively as part of my workflow for the last year. It has had time to become ordinary: a tool I use while doing my work, familiar enough that using it is no longer an event worth writing about.

Then I spent a week with Astra, and now I am writing about it.

GPT-6 Astra is a game changer for me. That is a strong phrase, but it is the honest description of how I feel after this week. I already liked working with Codex. Astra has made me excited about it again.

A year of regular use is useful context for that reaction. A coding assistant can be impressive the first time it does something you thought you would have to do yourself. The more interesting question comes later, once it has a place in your working day and you have some expectations of it. Does a new model still make enough of a difference that you notice?

For me, this one does.

There is also a less grand way to put it: Astra is a very nice model to work with. I mean that as a substantial compliment. When a tool is part of your routine, the experience of using it matters. You spend time in that interaction, and how it feels becomes part of how the work feels.

Technical discussions about models tend to give that quality less room than capability. I understand why; it is harder to put in a table. But I want both. I want help with difficult work, and I want to enjoy the process of getting there. My reaction to Astra includes that second part as much as the first.

I am a week in. I will have more to say as that turns into a month and the novelty settles into routine. For now, the thing worth recording is the reaction itself, while it is fresh: after a year of actively using Codex, Astra still managed to surprise me.

I am looking forward to week two.