Correct the request directly

no, I meant 7 x 8
what did I just ask?

Explicit corrections route the replacement request to the normal tools. Conversation recall uses the current session; it does not search someone else’s history.

Compare concepts and follow up

difference between precision and recall
give me an example
compare validation set and test set

Comparisons use authored definitions and selected distinctions. Unknown concepts are not invented. The configured local model remains the route for broader generated discussion.

Calculate several expressions

calculate each: 7*9
100/4
2/0

Up to eight arithmetic expressions can be evaluated independently. An invalid expression is reported on its own line.

Long prompts preserve the ending question

When local-model context is limited, the current request retains its beginning and ending with an explicit omission marker. Prompt truncation is reported in response metadata; this does not imply the model read every omitted sentence.

Learning is measured separately

The residual learner updates weights privately. Validation may reject a candidate. Stored source material remains unverified, and assistant output is not automatically fed back as training text.

Learning for everyone without publishing a chat

Shared learning is on by default. The chat header shows the current choice and offers Turn off; Options has the same setting. Previously saved off choices remain off. Future eligible user text becomes globally available as unverified source material and enters a separate shared neural training queue. Earlier private messages and assistant replies are not imported. Secret and personal-data filters are conservative exclusions, not a guarantee of anonymity. Do not submit private material while contribution is enabled.

shared learning status
shared learning off
shared learning withdraw
shared knowledge: your specific topic

Turning it off stops future contributions. Withdrawal removes retained contributions and clears the shared checkpoint for retraining from remaining text; other users may retain earlier replies. Private models never train from shared weights, so removing shared contributions does not require rewriting private models. Shared predictions are a fallback when no private checkpoint exists. The worker uses the same validation gates for both profiles. At most 256 shared chunks, 64 per contributor, are retained; each is at most 4,000 characters. Filters and quotas may skip messages. External LLM weights are unchanged.