The author compares AI-generated plans from Astra and Fable, finding Fable's output cleaner while noting Astra's overengineering tendencies and high quota consumption.
Adapted from @Voxyz_aiFor planning, I’m leaning toward Fable now. I’ve found that Astra still needs the rules I shared to keep Sol from overengineering. I’d add them to AGENTS.md. Recently, I’ve tried this on a few types of tasks, including writing and architecture planning for a new product. I had Astra and Fable each write a plan, then opened a fresh session with each model to compare the two. Both mostly favored Fable’s approach. I felt the same way reading the plans myself. Astra often adds overly strict gates to its plans. I think those requirements are worth discussing when choosing a plan, rather than getting halfway through the work and having to change rules that are blocking progress. Cross-model review can start at the planning stage. You don’t have to merge everything from both versions and keep making the scope bigger, haha. Also, Astra has been burning through my quota lately. I’m on the $200 plan, and I can use up more than half my quota in a day. I’ve just used one manual reset and have two left. Not sure how many days that’ll last. I’d save this back-and-forth planning and comparison for the more important tasks.
