these fable safeguards are fucking ridiculous...opus 5 is dog shit and fable is safeguarded to death....what exactly are we paying @AnthropicAI for????
My workflows now take 10x longer and cost 10x as much because I'm doing 5x the work just fixing its mistakes. And I can't even trust the fixes, because who knows what else it quietly broke in the background. Extremely disappointed.
Opus 5 has all the confidence of Fable with a fraction of the intelligence.
This model is a massive liability. It ignores narrow instructions, does things you didn't ask for, and stops when it shouldn't.
I have no idea what Anthropic did here, but it's a terrible product.