Many users criticized GPT-5.5 for leaving numerous bugs in builders' code, calling the model untrustworthy and warning against reliance on Anthropic releases.
No Digg Deeper questions have been answered for this story yet.
Most Activity
tricky little fucker
finding so many bugs it left behind with gpt 5.5..
@yacineMTB it is bad.
dont know how people have been so hypnotized with regards to any slop that anthorpic releases.
same repo different task. almost similar scope. copilot review found one low bug in codex's pr while 5 bugs( 3 high) in fable's pr.
@yacineMTB Can’t wait for GPT 5.6 to go head to head with Fable
@yacineMTB You can find them with fable too, just spawn another session - llms are stochastic
@yacineMTB Just wait till 5.6 there’s gonna be even more bugs found 😭
@yacineMTB question is who is lying ultimately
@yacineMTB so multiplied by the gpt 5.5 lies equals???/
@yacineMTB Same, I use gpt for double check things after claude models, and it always finds something. Even after Fable.
@yacineMTB have it run codex for code review.
i prefer going the other way normally, let 5.5 call claude when it needs to do interface design, etc.
but with fable i flipped it and it's been smooth
@yacineMTB I mean, to be honest, I don't clear anything until an external high tier model audits it.
Doesn't matter who makes it, it's gonna bounce.
@yacineMTB The real security threats were the bugs we vibe coded along the way.
@yacineMTB I did the same thing, so many bugs. You can do the same thing with opus. Although, doing this outside the anthropic ecosystem is probably the smart move.