LAM: When entering text, it will use UTF16/8 or something like that

Experimenting with the LAM to make the R1 create some HTML code.

For some reasons, the entered text is all weird until I realized that it is actually writing in UTF16/8 or something along that way.

All characters looks like they are enterered with 2 characters, thus creating a mess instead of writing the proper generated text.

See the screen capture.

EDIT: Just found out that it depends on the prompt. For ecample: Create a table for listing all startwars movies, will create this issue. But simply asking for a basic html file saying hello world, does not create that issue.

2 Likes

would love to see the prompts you use to make your requests.

1 Like

I did have a chat with Rabibit on X.com and while investigationg, i found out that it is not realted to the prompt itself.

Its a timing issue where entering text takes longer than expexted so R1 is assuming is failed and starts over. While the first thread is entering the text, a second and even a third one will overwrite the text creating an interlaced text entry.

The dev team is probably aware of the issue and a fix will be abailable soon.

Meanwhile, there is a quick workaround: just close the R1 for 1 or 2 seconds, it prevents the second text entry to start letting the first to complete.

1 Like