What counts as usage.

LADLE's usage cap is measured in tokens — the units the underlying model processes. Every message you send and every message LADLE sends back consumes tokens. There's no minute-based cap or query-count cap.

**What consumes the most tokens:** - **Long-context Projects**: attaching a 200-page PDF to a Project means that PDF is included in the context of every chat in that Project. If you have a chat back-and-forth of 20 messages, that PDF is processed 20 times. Big Projects are the single biggest usage driver for most heavy users. - **File attachments in chat**: attaching a 40-page file to a single message uses the tokens for that file once. Same file in the next message in the same chat: uses them again. - **Long conversation histories**: LADLE reads the whole conversation on every message. A 100-message chat uses more tokens per new message than a 10-message chat, because the history grows.

**What consumes less than you'd think:** - Short back-and-forths, even many of them. 50 quick "what's the capital of X" queries barely dents your cap. - Streaming responses — the counted tokens are output tokens, not "time streaming". - Voice input via the speech-to-text feature (transcription happens locally in the browser; only the resulting text is sent).

**What doesn't count at all:** - Time spent reading previous chats. - Time spent in the app not sending anything. - Failed messages that error out before reaching the model. - Chat exports and account settings changes.

**Where to see your usage:** Settings → Usage shows a rolling 30-day chart of tokens consumed, broken down by category (chat, Projects, files). It also shows the current tier's cap and how close you are to hitting it. The chart updates every ~10 minutes.

**Why we don't hide this**: some AI products treat "usage" as a vague opaque metric. We show you the actual number because you're paying for it — and because the pattern of "how does this get expensive" is worth understanding whether you're on LADLE or any AI subscription.