Anas

Vibing

[Anas @Anasvtqf]

Just fork the thread and test it

[imjustnewatai @imjustnewatai]

One thing they aren’t telling you about OpenAi 5.6 is that it doesn’t work well with older models like GPT-5.5. It’s recommended that you don’t use 5.6 on an already-existing 5.5 Codex thread. Start a fresh thread instead.

[Anas @Anasvtqf]

I am absolutely hammering Devin’s GLM 5.2 and SWE 1.7. $20 is a steal

[Anas @Anasvtqf]

A lot of the startup space is just grifting and larp , when you don't do that then you don't get far. So it's like wtf...

[Anas @Anasvtqf]

@cherry_mx_reds Real asf bro

[Anas @Anasvtqf]

@usr_bin_roygbiv Do you think this gap of knowledge in Ai stuff compounds over the years or LLM's become good enough to close the gaps themselves?

[Anas @Anasvtqf]

@wolfie_ Gotta always brush up

[Anas @Anasvtqf]

@wolfie_ Seems peaceful

[Anas @Anasvtqf]

I just had a dream that i had 5.6 sol low/no thinking on cerebras clocking 1000 tok/s...

[Anas @Anasvtqf]

@usr_bin_roygbiv The way 5.4 xhigh understood intent was magical, alongside the fact it solved compaction. One monothread would cook for ages.

[Anas @Anasvtqf]

@ChrisRinvesting Fair enough that is true tbh

[Anas @Anasvtqf]

Final verdict: For big, long horizon tasks and refactors = OMP all day, for little quick stuff = Codex App.

[Anas @Anasvtqf]

@ibocodes but self hosting on aws and etc beats it

[Anas @Anasvtqf]

@ChrisRinvesting For the little stuff, it does too much like making a big to do list, spawning subagent and random bs when I literally want one line changed lmao

[Anas @Anasvtqf]

Just give me 5.6 man

[Anas @Anasvtqf]

@alexabelonix W bot. Give me your api keys here.

[Anas @Anasvtqf]

Ai psychosis is an IQ test, spend 10 mins speaking to any LLM and you'll know what I'm talking about. The same repeated thought patterns.

[Anas @Anasvtqf]

@argofowl gpt still better for raw engineering and grunt work, fable is just nice to talk to and understands user intent well

[Anas @Anasvtqf]

@usr_bin_roygbiv Excited for what's to come !

[Anas @Anasvtqf]

@LexnLin I reckon it's just reg 5.5 limits, just compared to fable is an insane amount bc it's expensive and inefficient.