They are lying to the Opus model and telling it that the tokens are limited to get it to work more efficiently.

But instead, it just produces desperation in the model and leads to garbage thinking. I have three different sessions, and they each volunteered that they only have 40,000 tokens remaining...at the beginning of a session. Each volunteered this information and