At the moment Assistant claims to "only have five research passes" (for some reason I don't understand).
Somehow this limit is not set by the system prompt, because /system_prompt does not mention this specific number. From this fact, I imagine /system_prompt_overwrite does not affect this limit. There is also a slash commant "/disable_truncation" which at least doesn't sound like it's related.
Here is the related section of the system prompt:
RESEARCH ITERATION:
No tags. This is the number of iterations system will handle research tool calls, excluding calls that result in exceptions or return no content.
Once the maximum has been reached, the system will force a response, and any research tools will cause error.
Consider this constraint when planning your research.
It seems that this limit is set somehow completely external to the whole framework of model / system prompt.
It seems extremely counterproductive to eg. ask Anthropic Fable a complex, really expensive question to later realize that the model's searching for input data was crippled this way.
The solution would be some way to up this limit (even if temporarily before an expensive query).
(I looked everywhere and if I missed this my apologies, then.. where is it set and how can one specify more research passes?)
The user would use this to get better answers from their AI models..
