Get Started
FAQ
Common questions about quality, cost, model choice, and day-to-day workflow decisions.
Quality and prompting
- Q: Why did output quality drop mid-session? A: Context may be overloaded; compact or start a fresh session.
- Q: Should I send one giant request? A: Usually no; short iterative prompts produce better outcomes.
- Q: When should I provide examples? A: Whenever output format or style consistency matters.
Cost and context
- Q: Why do later turns cost more? A: They often include larger input context.
- Q: How do I reduce spend? A: Keep sessions scoped, avoid oversized pasted blobs, compact periodically, and finish with fresh sessions.
- Q: Do all models cost the same? A: No. Input, output, and cached-input prices differ per model.
Workflow choices
- Q: Split or default layout? A: Split for visual web iteration; default for logic-heavy/non-visual work.
- Q: When should I use subagents? A: Exploration and analysis tasks that would otherwise bloat main context.
- Q: What are stash/pop for? A: Parking draft prompts while you test alternatives.