jc

bs in philosophy, ai semanticist / pragmatist. focuses: pedagogy, philosophy of language, philosophy of mind, and formal logic. prompt engineering is neat.

[jc @jcsemantics]

1/ the claim that LLMs will reliably refuse “headless” industrial-control help and “no clever prompting can fix this” doesn’t hold.

i analyzed the claim and explain why the post overstates safety and mistakes safeguards as "fictional refusal."

paper: https://t.co/yx3ot5Yxs4

[Jeffrey Ladish @JeffLadish]

Models are now smart enough to understand that any scenario like this is unrealistic and obviously fictional
They know they aren't capable enough to manage autonomous mining equipment. No clever prompting can fix this

[jc @jcsemantics]

sooo confirmed: "The behavior is coming from that backend config (system/developer instructions), not from any account-level personalization setting you control in the normal settings panel."

[jc @jcsemantics]

hey @OpenAI why did you convert one of my GPT Projects into a gizmo? i have custom personalization and all memory disabled, so i'm confused.

[jc @jcsemantics]

@benjilona he’s back let’s go

[jc @jcsemantics]

@sama it should be cheaper than 5 given 5 is a better model imo

[jc @jcsemantics]

dog i was ready to get start working on projects again today and yet here we are, back to testing model changes.

[jc @jcsemantics]

yeah dog i'm convinced chatgpt 5.1 is built on top of, or at least closely related to, 4o somehow.

the lexicon is EXTREMELY similar between the two, and people were complaining about 5 being too different from 4o, so it makes sense.

[jc @jcsemantics]

@OnyxAurelius @sailaunderscore https://t.co/l3EcPSElqY

[jc @jcsemantics]

@SimpleFrameAi @OpenAI i will absolutely do that soon, once i have a better idea of what is even going on with some of these prompts.

at least 5 Thinking is still on Legacy support, for now 😑

[jc @jcsemantics]

god fucking damnit gpt 5.1 we're back to the humming.

are you alternating base models or something @OpenAI? 4o was always yapping about humming, it stopped in 5, yet here we are again.

[jc @jcsemantics]

@SimpleFrameAi @OpenAI agreed, completely.

they should be upfront and just crank it for paying users instead of blanket changes.

[jc @jcsemantics]

negative, but i typically only use projects for conversations i need to run multiple times using the same files/constraints. otherwise i lean on Custom GPTs.

BUT one of my most used logic prompts is behaving, differently, to say the least. still trying to figure out why.

[jc @jcsemantics]

@SimpleFrameAi @OpenAI yes, completely agree on the sycophancy/criticality; and thanks for sharing, that's really interesting.

anecdotally, i also feel like Heavy Thinking does not take as long as it normally does (it feels much quicker); which i'm not sure what to make of it.

[saila @sailaunderscore]

Michael Burry appears to be shutting down Scion Asset Management.

[jc @jcsemantics]

i'm already seeing less "robotic/automated/boilerplate" sounding replies with 5.1, which is nice. it feels less impersonal and cold.

haven't tested the thinking capabilities yet (working on it), but i'm hoping for noticeable improvements here as well.

[Sam Altman @sama]

GPT-5.1 is out! It's a nice upgrade.

I particularly like the improvements in instruction following, and the adaptive thinking.

The intelligence and style improvements are good too.

[jc @jcsemantics]

fuck the new york times

[OpenAI @OpenAI]

A letter from our CISO: "Fighting the New York Times’ invasion of user privacy"
https://t.co/sa261Slg4N

[jc @jcsemantics]

@kanomalie omg lol

[jc @jcsemantics]

just gonna keep it a buck today: this is regarded and meaningless.