My honest experience with Astra.
- 发布
- 09/06 11:53
- 收录
- 09/06 18:00
- 来源类型
- 开发者社区
- 档位
- 社区
- 信源状态
- 正常
Just to be clear, I’m on Pro 20x.
When Astra dropped in the Mac desktop app, I was sitting at 39% of my weekly usage with 2 banked resets.
In the ~3.5 hours since getting access, I’ve run 6 Astra Ultras,roughly 20 minutes each. One was a full repo audit, and the other five were deeper dives into specific parts of the codebase.
I also ran 1 Astra Max to actually implement the findings from those audits.
And right now I have 3 more Astra Ultras running concurrently: two implementation/coding tasks and one basically red-teaming the work Astra Max just did.
I’m fully aware this is probably not the most token-efficient way to use it tho i got a save token skill on Taku that saved 12% usage.
But I honestly don’t care right now because it’s fixing things Sol has spent weeks overengineering.
It’s finding gaps I missed, pointing out where the architecture went sideways, simplifying things that became unnecessarily complicated, and actually pushing the repo instead making extra troubles.
As of right now, I’m down from 39% to 26% weekly usage, still with 2 banked resets remaining which is pretty much the only reason I’m comfortable abusing Astra Ultra this aggressively. Thanks Tibo.
Again, I understand the burn rate.
But so far, I’m happy with Astra. For the kind of repo-level work I’m doing, it feels significantly better than Sol at actually getting things done.