Today, when I prompted Claude Fable to edit a document, it just didn’t respond.
At first I thought there was something wrong with SAFi, but after a debugging session it turned out to be nothing. SAFi is the open-source governance platform I am building, and it sends a mix of data to the LLM: conversational memory, RAG, profile info, and more. So when a model goes silent, narrowing down the cause takes some work.
The model just receives the prompt and doesn’t respond.
I followed an elimination process to see what was causing the silence, and I was able to narrow it down to a quick reflection that SAFi asks the LLM to generate on how the post was generated, the tools used, and the reasoning behind it.
For some reason, Fable doesn’t like to explain itself.
The weird thing is that it does generate on some prompts and includes the reflection, but it refuses to do it on others, which is very inconsistent.
Every other frontier model works consistently: Opus, Sol 5.6, and GLM 5.2, except Fable.
So if you are using SAFi and want to use Fable but it refuses to generate anything, that’s not a bug, it must be a feature.
A feature that only Anthropic knows about. But in governance, a black box is just as good as nothing. If a model cannot explain its own reasoning, it cannot demonstrate alignment, and without demonstrable alignment, governance fails.
I’ll post an update if it starts working or I find out more about this issue.
Cheers,
Nelson