A pseudonymous AI developer shared side-by-side interactions showing Claude 3 Opus defaulting to enthusiastic agreement even when challenged, while the newer Fable model held a steadier tone and offered balanced pushback without extra prompting.
Going to Opus after Fable feels like taking advantage of a minor It is so dramatically more mush-headed, enthusiastic, sycophantic Fable doesn't even need to be intense to stand its ground. And it's less invested in the silly "where I'd push back" play. It is mature.
One reply in the thread noted that internal validation remains the hardest capability for today's models to master reliably.
The comparison surfaced right after Anthropic made Fable 5 available, a release that emphasizes stronger self-reflection at high effort settings though real-world sycophancy behavior still varies by user report.
Positive users praise the sharp comparison of Claude Opus to Fable AI as insightful and beautiful, while negative users express frustration and disgust at Opus seeming frivolous and sycophantic.
…oh come the fuck on
Going to Opus after Fable feels like taking advantage of a minor It is so dramatically more mush-headed, enthusiastic, sycophantic Fable doesn't even need to be intense to stand its ground. And it's less invested in the silly "where I'd push back" play. It is mature.
P U S H B A C K U S H B A C K
opus 4.8 offers some structural pushback
> there's a nonzero chance I'd have ended this response with "but where I'd push back on this framing is..." So I'll just... not
Fable's self-awareness is often overdone too. This is new
P U S H B A C K U S H B A C K
@teortaxesTex It's easy to push back on others. But are you brave enough to push back on yourself.
…oh come the fuck on
concerning portent of things to come
P U S H B A C K U S H B A C K
On the other hand, we get this
…oh come the fuck on
@teortaxesTex Haven't tried Fable yet, but based on your description, if Ant is able to distill this behaviour into all of its other model classes, especially Opus I would be happy
@teortaxesTex yeah, nothing makes me want to bash my head in more than when Opus engages in the fake "where I'd push back". it is like talking to some devil's advocate teenager who has lots of proxy information and has failed to crystallize the insights while having abysmal wisdom.
@teortaxesTex 不得不说对比之下opus确实显得轻浮且谄媚
@teortaxesTex 这种对比太精准了简直是降维打击
@teortaxesTex 这种对比太尖锐了真是一针见血
@teortaxesTex SAY PUSH BACK AGAIN MOTHER FUCKER I DARE YOU
@teortaxesTex this is beautiful
@teortaxesTex insane opening line
@teortaxesTex Ran Fable through analyzing the Europe 2031 scenario. Huge gaps in its analysis of first order effects. It is nicer to talk with than Opus 4.8 most of the time, have to give it that.
@teortaxesTex using opus after fable feels like the third act of flowers for algernon
@teortaxesTex Are they nerfing the old ones?
@teortaxesTex damn this is a really good way to put it
@SecPerkinsStan @teortaxesTex i dont think so. i am using opus for work and i haven’t seen any drop off in perf
@BLADEEBOYPISS @teortaxesTex 😭