Review types
Pick the review that matches the job. Review type is how much of the set to cover. Thinking is how hard Assistant reasons. Cost is about $0.50/sheet either way — Comprehensive just covers more sheets and disciplines.
| Job | Review | Thinking | Typical cost |
|---|---|---|---|
| Spot check — a few sheets, one discipline | Light | Low | ~$0.50/sheet on a sample of ~5–10 sheets |
| Full set before you send a package | Thorough | Low, or Medium if coordination is messy | ~$0.50/sheet |
| Every discipline, high-stakes QA | Comprehensive | Extra High | ~$0.50/sheet. Warns and waits. |
- Light — a sample, a handful of issue types. Cheap enough for everyday QA.
- Thorough — the full in-scope set, focused issue types. Drops low-yield checks.
- Comprehensive — every in-scope discipline. Does not skip checks to save money.
All three write the same findings table and markups.
How to run each
Spot check. Attach the sheets. Ask for a review. Pick Light. Leave thinking on Auto or Low. Steer with follow-ups — "only flag what a GC would raise as an RFI" or "go deeper on the door schedule."
Before a send. Same start. Pick Thorough. Use Medium if the set is messy.
Every discipline. Pick Extra High thinking (for example Sonnet 4.6 Extra High). Attach the full package. Ask for a comprehensive review of every discipline. Assistant surveys the set, shows a cost picture, and waits. Confirm only if you mean to spend.
Turn a repeatable gate into a workflow with Extra High thinking and Comprehensive in the instructions.