yessir

[yessir @JanahYsr]

@petergostev what is the difference between the two? (in the context of LLMs)

[yessir @JanahYsr]

@ForwardEditor the only time it really backfired is when @mattshumer_ got his Home folder nuked and they did bunch of posts/usage resets/etc ..
x / reddit complaints from randos don’t count for them

[yessir @JanahYsr]

@badlogicgames I totally agree - as models get “better” do you think specialized domain knowledge would be as beneficial when steering LLMs?

[yessir @JanahYsr]

@kimmonismus Opus was good when internal employees were **using** it. This ceased to be true since Mythos.

[yessir @JanahYsr]

@Angaisb_ well if you can afford distributing your work accross the 5 hour limits then it should be ok

[yessir @JanahYsr]

@athyuttamre @jxnlco doesnt interact well with codex tasks

[yessir @JanahYsr]

@LexnLin nothing ever happens

[yessir @JanahYsr]

@krzyzanowskim same “Slop” app for 2000$ each

[yessir @JanahYsr]

@figma @jsngr wassup

[yessir @JanahYsr]

@Kreshomir @jsngr hello

[yessir @JanahYsr]

@MarcosHernanz this is shocking

[yessir @JanahYsr]

@jxnlco ChatGPT Pro model does not have access to local files. again, what is even the point of having a desktop app

[yessir @JanahYsr]

@jxnlco no study mode / deep research in the desktop app. what is even the point

[yessir @JanahYsr]

@alxfazio true, although most magic happens in server side services (e.g. codex auto-compact)

[yessir @JanahYsr]

@LexnLin there is no reason for them to come anymore

[yessir @JanahYsr]

@Hi_Mrinal maybe worth a try - I have noticed the last gen of models are so RL fried that it is impossible to make them speak like humans, let alone making them teach you. but I had some decent success with chatgpt study mode

[yessir @JanahYsr]

@EvaXYZH bad for the token business