Having Camelia draft a high-quality deliverable with the review loop

For big analytical deliverables, Camelia uses a self-improving review loop instead of a single draft. It's designed for substantial pieces — market analyses, competitive landscapes, business-plan sections, product specs, go-to-market plans, viability studies — that benefit from grounded research plus iteration.

How the loop works

First, Camelia drafts the deliverable grounded in live web research. Then a separate reviewer — "Camelia-B," which shares none of the drafter's context — scores the draft from 0 to 100 and critiques it. The drafter then refines based on that critique. This drafts-then-reviews cycle repeats for up to three refinement passes (capped at four total), stopping early when the reviewer marks it done or the score reaches at least 80.

Why the separate reviewer matters

Because the reviewer starts fresh with no shared context, it evaluates the work the way a new reader would — catching gaps the drafter might rationalize away. That independent critique is what pushes quality up across passes, rather than the model simply agreeing with itself.

Where the result lands

The finished deliverable is saved to your project's Documents folder, so it becomes a real project artifact you can open, edit, and reuse — not just a chat message.

When to use it

Reach for the review loop when you want a polished, research-backed document. It is not meant for quick answers or simple file writes — those are faster handled by a direct response or a single sub-agent. Tell Camelia the deliverable you want (for example, "a competitive analysis of shipping-container housing startups with pricing"), and she'll run the loop and save the result.

Expect this to take longer than a normal reply, since it involves research and several review passes — that extra time is the point.