I am using Lynx AI on the LingQ website while studying Chinese. The first message generated in a new conversation correctly reflects my saved word statuses: words I know appear in the Lynx-generated text as known, while unfamiliar words receive the appropriate highlighting or pinyin.
However, after I reply and Lynx generates its next message, many words that I already know are displayed as though they are unfamiliar. This appears to happen consistently after one or more exchanges. The underlying text is fine; the problem seems to be that later Lynx messages are not using, or are not correctly loading, my saved vocabulary statuses.
Steps to reproduce:
- Start a new Lynx AI conversation.
- Inspect the word statuses in Lynx’s first generated message.
- Reply to that message.
- Inspect Lynx’s next generated message.
- (It might take a couple of such exchanges to reproduce the issue)
Expected result: Every generated message should use current saved word statuses consistently.
Actual result: The first message displays them correctly, but subsequent messages mark many known words as unfamiliar.
A possible clue: When Lynx generates the first response, the word statuses initially appear incorrectly for a brief moment, but then automatically refresh to match the saved vocabulary statuses. For the second and subsequent responses, this refresh never occurs, so the initial incorrect statuses remain displayed. It appears to me that the normal post-loading status update is only being applied to the first response.
In the screenshot below, the first Lynx response is good and the second reflects word status incorrectly.
