Users praise GPT-5.5 for its improving bug-fixing and coding skills because they see it outperforming Fable and eagerly await versions like 5.6.
No Digg Deeper questions have been answered for this story yet.
Most Activity
what
gpt 5.5 is finding and fixing bugs that fable left behind
like really fucked up bugs. weird morton z order space filling curve bugs
@yacineMTB Use the Triadic model. That way you get the power of fable five, GPT 5.5, and super Grok…. IT is exponential. And believe me, when the three of them are together, they work better than the three of them alone.
what the fuck
@whatdafuqkyle @yacineMTB with AGY I generally still to either the Flash Medium or Gemini 3.1 Pro. I still lean alot toward @goose_oss ; powerful combo with Deepseek and Qwen-coder.
@FroidEtCold @yacineMTB antigravity with opus 4.6 was juicy for me tn.
been a while since I used it.
its actually totally normal that even the most advanced models mess up sometimes. another cool thing is sending your code or ideas for a review to a completely different model family. i made a skill that lets one model call another harness for a review. like claude calls codex and opencode (glm-5.2/kimi2.7). it works insanely well, especially when all models agree (though even that is not a 100% guarantee, but still)
@yacineMTB It has bee doing that since Opus 4.6 started acting up in May. I would run codex when I disagreed with Claude code and it would prove me right + find a ton of other thing.
Codex has just been super impressive this year...
@x_micky_b @yacineMTB doesn't it over complicates? compared to GPT 5.5 Medium
@yacineMTB Fable 5 for planning, GPT 5.5 for coding, Fable for review/QA/testing...
Give us 5.6 🙏
@yacineMTB Fable is really annoyed with Codex, yep.
@yacineMTB 5.5 my GOAT, can't wait for 5.6
came up with a dumb idea to filter text models before you press enter. apparently that doesn’t always happen for most end user critical mass models.
most of the crap I’ve vibe coded has looked much worse than this, and it appears to work.
could be open source tool for users to use before promoting, or lab level idk.
@yacineMTB Got has always been a great reviewer but is really bad to interact with as well as writing something from scratch
@yacineMTB I tried Fakel, was disappointed. I am going to stick to GLM 5.2 + GPT 5.5. Better overall, and doesn't cost an arm and a leg.
@yacineMTB I know. Had to go through that during the first Fable release already.
@yacineMTB The inverse is true also
@yacineMTB Yeah gpt 5.5 is so slept on
@yacineMTB And vice versa for me. This is why using them both together works so well.
@yacineMTB As Tibo would say: