Not GP, but personally I don’t trust claude. Work does for some reason, and work strongly encourages using it, but i stay far away outside of that.
I think every product in this space should have to clear a very high bar for trust rather than making it a race to the bottom with whataboutisms.
Anthropic use Deepseek behind the scenes instead of their own models, whenever they feel like it. People have shared outputs from Claude that show internal tool errors the models have accidently returned with Deepseek error codes.
I'm curious about this, where'd you learn that?
There were quite a few people from the AI community on X sharing screenshots.