Another Astra Post
- Published
- 09/06, 12:43
- Ingested
- 09/06, 15:00
- Source type
- Dev community
- Tier
- Community
- Source status
- Healthy
Ok, so I have been hard Claude Code from the beginning, and for context, I’m an actual software developer, not a vibe coder. I use Claude Code daily both personally and professionally.
So I took the leap to try Astra 6 and signed up for a Pro 5x plan over there. Coming in blind, the codex app for Mac is very similar to the Claude Code app for Mac, and it imported my chat history from Claude,which meant it had relatively good context for all the projects I was working on. I liked that.
I have been going around and around with Fable for weeks on a bug we just can’t reliably squash on one of my personal projects. I thought I would give Astra a crack at it.
So first things first, it seems less verbose than Claude overall, which I like. It also was able to read Claude’s indexed memory about my project, which gave me confidence it knew what was up straight away.
Overall though, we went round and round through multiple turns trying to get it to properly understand what we were even focusing on. I ultimately went back to Fable and had it write me a handoff prompt for Astra to orient it, it did, Astra still seemed unsure. It did not feel confident enough in itself to even do a design review on a bug fix.
So I’m not sure how I feel. On the one hand, Claude couldn’t squash my bug in multiple iterations, but on the other hand, it seems like Astra refused to try.
I feel like at least with Claude, sometimes we are throwing spaghetti at the wall and hoping it sticks, and sometimes we are over engineering something that simpler logic and proper steering would fix, but it still feels like progress. Astra felt like a lot of effort with no action plan.
So I turned it to a new feature I’ve been thinking about because if Claude couldn’t fix my bug, I guess I shouldn’t expect a miracle from Astra. For the new feature, it did design and implement relatively easy, and it worked easily, but wouldn’t say better. As good, maybe slightly better given usage constraints? I think that’s about the only place Astra is going to edge out Fable here. I did all the big investigation turns and a whole design review, build, and adversarial refute session on 5x on Astra and barely made a dent in my usage.
TL;DR - I think Astra is as good as Fable, maybe even marginally better, but in my opinion this is NOT a generational leap and certainly not AGI if that’s even a thing we could define.