Is it me or did Codex 5.5 suddenly get better in the last 24 hours 👀

@dreamsofcode_io i haven't noticed anything

@dreamsofcode_io idc im too busy using fable

@dreamsofcode_io Idk if i am psychotic, but after doing a basic lora finetuning a few months ago, i understood a model’s “nerf/boost” settings can pretty much be “hotfixes” which makes all this nerf/boosting super possible Look at what sampling settings are ref: denislavgavrilov.com/p/actionable-l…

@dreamsofcode_io marginlab.ai/trackers/codex…

@dreamsofcode_io probably just so many fewer ppl using it today 😂

@dreamsofcode_io i think it went back to baseline. last week was pretty bad

@dreamsofcode_io just you

@dreamsofcode_io I noticed this! So glad I’m not the only one. The “people are making direct comparisons and running benchmarks” knob has clearly been clocked up.

@dreamsofcode_io it does feel great today, and over the last few days

@dreamsofcode_io Maybe, but I’ve literally just been getting fable to come up with the instructions and then 5.5 to execute those It has been kind of a crazy hack to not hit the fable limits

@dreamsofcode_io No it is the same. I was using it with CC recently and did not notice any changes while updating it every time it asked.

@dreamsofcode_io It's you

@dreamsofcode_io no, it's LSD

@dreamsofcode_io Memory? I haven't noticed a difference, still using it heavy.

@dreamsofcode_io because anthropic is a retard

@dreamsofcode_io prompt it with fable for max degeneracy and progress