What happens when you hit the cap.

When you hit your plan's rolling usage cap, LADLE doesn't cut you off — it throttles new-message send-rates until your rolling window recovers. In practice: you can still chat, but there's a pause between messages (typically 30-90 seconds) instead of the normal near-instant send.

The rolling window is 24 hours. Usage from 24 hours ago drops off; usage from an hour ago is still fully counted. Most people who hit the cap once find that after ~4-6 hours of normal use, their rolling total drops back below the threshold and the throttle lifts on its own.

The pause is a real pause, not a "degraded model". You get the same Claude Sonnet response — you just wait a bit for it. We chose this over the alternative pattern (hard cutoff or quality downgrade) because "wait a moment" is honest; "your response is worse now" would be dishonest.

**If throttling is happening regularly, three moves:**

1. **Check what's driving usage** in Settings → Usage. If one Project accounts for 60% of your usage, the fix is to reduce the context attached to that Project (do you need all 200 pages, or would 40 pages plus a summary work?). 2. **Split long chats.** A chat with 150 messages consumes disproportionately more per-message than five 30-message chats on the same topic. If you're in a genuinely long-running project, a Project with focused sub-chats is more efficient than one giant thread. 3. **Consider Max 5x.** If your usage-driving pattern is legitimate work that you can't compress, Max 5x is likely the right answer. See "Base vs Max 5x vs Max 20x". This is not us pushing — it's often actually the correct move for heavy legitimate use.

**What we won't do:** stop your chats mid-response, cut off access without warning, or downgrade the model quality quietly. If throttling starts, you'll see a small notice in the composer explaining what's happening and where to check your Usage dashboard.