[Bug] Fable model incorrectly triggers safeguards on scientific analysis discussion
Open 💬 0 comments Opened Jul 8, 2026 by mikegilchrist
Bug Description
I have just tied up a complex, but well principled, data analysis of bird song data using sonnet (should have used opus in retrospect). I wanted to use fable to organize the framing of the analysis and the results. However, after my first message I hit Fable's safeguards and got booted to opus.
Here's my message:
We have now finalized our analysis. Here are *some* key messages we want to emphasize in our presentation of our results.
1. The data depends on not just variation in behavioral responses between males, but also variation in responses within a male (the 'shape' parameter in the NB2 fitttings and what I referred to earler as male inter-trial variation but would better described as something like male repeatability, male reliablity, male consistency, ...). The only way we can get a handle on this inter- and intra-male variation is via repeated experiments with the same male. THis is a key insight and would mean that each male should be tested repeatedly instead of just once. This would greatly improve the efficiency of animal usage, to boot.
(to be continued in the next message)
It seems quite an over-reaction to have triggered safeguards and be booted back to opus.
Environment Info
- Platform: linux
- Terminal: tmux
- Version: 2.1.198
- Feedback ID: 5084d088-b0f1-43d9-9802-805ad0a6df6a
Errors
[]