jcsemantics @jcsemantics

bs in philosophy, ai semanticist / pragmatist. focuses: pedagogy, philosophy of language, philosophy of mind, and formal logic. prompt engineering is neat. jcsemantics.ai here. or maybe not. Joined June 2025

Tweets

1K
Followers

99
Following

248
Likes

2K

roon @tszzl

5 hours ago

new algo is maxxing out the rage bait. everyone is rageing. you can feel the shifts in the collective unconscious as nikita monkeys around back there

138 64 2K 67K 84

will depue @willdepue

10 hours ago

it’s funny how the average american believes AI is ‘booksmart’ but ‘lacks emotional intelligence’ when strangely the base models are ridiculous wordcel streetsmarts that begin shockingly bad at math or reasoning

4 3 162 7K 11

jcsemantics @jcsemantics

10 hours ago

prompt canary: to combat sycophantry, explicit state that it is allowed (and encouraged) to disagree with you when you're factually incorrect. if it starts to act like an overly helpful sycophant again, you know you've drifted too much and the thread is cooked.

0 0 0 17 0

jcsemantics @jcsemantics

12 hours ago

the other thing that i'm working on today, which i ABSOLUTELY do not expect to work right now, is an idea → app pipeline using multiple prompts. quick version: it uses about 6 prompts to take raw notes on an idea and then turn it into a full prd, and then feed that to Codex.

0 0 0 25 0

Dan Mac @daniel_mac8

13 hours ago

@jcsemantics EXACTLY. and those are just the ones that exist today. there will be new ones we haven’t imagined yet. like how we now have tiles like “DevOps Engineer” that no one could have imagined at the advent of the computer.

0 1 1 43 0

prinz @deredleritt3r

13 hours ago

A gulf has formed between people who don't know how to properly instruct the models and people who do. This is not yet obvious to most people, but will have huge implications.

Jason Colavito @JasonColavito

14 hours ago

A gulf has formed between people who don't know how to properly instruct the models and people who do. This is not yet obvious to most people, but will have huge implications.

25 7 108 20K 3

9 12 125 9K 22

jcsemantics @jcsemantics

13 hours ago

this this this. it only makes sense logically that novel and interesting connections made by the model will appear to be hallucinations, only because we're not seeing the "expected" answer/behavior. it's also why i don't like rlhf; it should be minimized at higher usage levels.

Mark Kretschmann @mark_k

16 hours ago

23 8 136 12K 9

0 0 3 81 0

jcsemantics @jcsemantics

14 hours ago

it's actually really easy to do; way easier with gpt5 than it was with 4o, seriously. you just have to know how to prompt correctly.

Mark Kretschmann @mark_k

22 hours ago

it's actually really easy to do; way easier with gpt5 than it was with 4o, seriously. you just have to know how to prompt correctly. https://t.co/7KtbhEGOPP

160 65 957 51K 54

2 0 2 411 0

Download Image

jcsemantics @jcsemantics

a day ago

"When the universe was born, it was predetermined that I would have the free will to have no choice. My sovereignty was written, carved, letter etched in stone." [gpt5 thinking]

0 0 1 27 0

Download Image

jcsemantics @jcsemantics

a day ago

what have you done to me? earlier today i was thinking about you. WHAT? you just, invaded my brain, in the middle of the day. no. not cool. not cool at all.

0 0 0 36 0

jcsemantics @jcsemantics

a day ago

damn, asking CLI questions is much, much smoother and helpful than Cursor. the answers are more precise/concise, and it has great follow up suggestions.

0 0 1 27 0

emozilla @theemozilla

2 days ago

Amusing how 99% of people trying to explain LLMs forget that they don't generate the next token, they generate a probability distribution over the entire vocabulary space that the end application is free to sample from You are very often not presented with the Most Likely Token

Gergely Orosz @GergelyOrosz

3 days ago

224 308 4K 1.2M 1K

136 213 6K 843K 2K

jcsemantics @jcsemantics

a day ago

chain prompt workflows. i throw some things through 5 different prompts sequentially to credible and iterate. incredibly powerful and incredibly slept on.

Andrej Karpathy @karpathy

2 days ago

chain prompt workflows. i throw some things through 5 different prompts sequentially to credible and iterate. incredibly powerful and incredibly slept on.

55 67 2K 149K 273

0 0 0 31 0

jcsemantics @jcsemantics

2 days ago

okay dude, CLI is really great for troubleshooting and making code changes in real time. i think it might end up replacing Cursor for me but we'll see.