This prompted the question about the data source for Mythos/Fable training. The security bugs reports, fixes are somewhat of a sensitive private collections.
I did check the documentation from Anthropic on this, it only mentions the usage of publicly accessible data.
Any ideas how Anthropic would have got their hands on this private data if they used that at all?
filup•1h ago
The benefit with the models is that you can get rid of alot of the grunt work and just try a massive amount of things that might work all at once in a short amount of time in a automated fashion.