\
  The most prestigious law school admissions discussion board in the world.
BackRefresh Options Favorite

Sam Altman: maybe selling AI by tokens was a mistake (link)

https://x.com/businessinsider/status/2062211094450434219?s=6...
disrespectful marketing idea
  06/04/26
american closed model providers are going to get their assho...
razzle meetinghouse
  06/04/26
no corp is gonna run chinese crap at enterprise scale
Histrionic swollen old irish cottage
  06/04/26
...
diverse ticket booth kitty
  06/04/26
...
multi-colored box office
  06/04/26
Everything else is made in China. China provoded tons of ser...
Thirsty judgmental location french chef
  06/04/26
its more comparable to telecom prohibitions than importing r...
Histrionic swollen old irish cottage
  06/04/26
Why are the telecom prohibitons necessary if corp wouldn't u...
Thirsty judgmental location french chef
  06/04/26
the US government will just make it illegal to use chinese A...
aggressive brass community account
  06/04/26
They are all distilled counterfeit garbage anyway
diverse ticket booth kitty
  06/04/26
...
Histrionic swollen old irish cottage
  06/04/26
yeah but he's right that a lot of companies would otherwise ...
aggressive brass community account
  06/04/26
enterprise foundation models dont charge per token except fo...
Histrionic swollen old irish cottage
  06/04/26
They can just use cheap US models. What jobs really need fro...
diverse ticket booth kitty
  06/04/26
they can't and won't
razzle meetinghouse
  06/04/26
ofc they can, the whole US economy is going to revolve aroun...
aggressive brass community account
  06/04/26
...
Histrionic swollen old irish cottage
  06/04/26
They can and won't.
Thirsty judgmental location french chef
  06/04/26
no, they can't. we are completely beholden to china wrt trad...
razzle meetinghouse
  06/04/26
This is all reasoning for the "won't" bucket. We c...
Thirsty judgmental location french chef
  06/04/26
has anyone tried coding with chinese models?
Obsidian Marvelous Laser Beams University
  06/04/26
Yeah it's not horrible but now that codex 5.5 is insane you ...
diverse ticket booth kitty
  06/04/26
is it really that much cheaper or better than claude code?
Obsidian Marvelous Laser Beams University
  06/04/26
Yeah it's significantly cheaper Open AI appears to be heavil...
diverse ticket booth kitty
  06/04/26
They nerfed codex usage limits massively on the 1st
Orange misanthropic church skinny woman
  06/04/26
How many times have people said "Have you ever tried us...
Thirsty judgmental location french chef
  06/04/26
I don't have coding needs (until I try to make a video game)...
Stimulating drab ape
  06/04/26


Poast new message in this thread



Reply Favorite

Date: June 4th, 2026 11:21 AM
Author: disrespectful marketing idea

https://x.com/businessinsider/status/2062211094450434219?s=61

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914690)



Reply Favorite

Date: June 4th, 2026 11:28 AM
Author: razzle meetinghouse

american closed model providers are going to get their assholds eaten out by low cost chinese models and are scared shitless

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914700)



Reply Favorite

Date: June 4th, 2026 11:28 AM
Author: Histrionic swollen old irish cottage

no corp is gonna run chinese crap at enterprise scale

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914701)



Reply Favorite

Date: June 4th, 2026 11:35 AM
Author: diverse ticket booth kitty



(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914719)



Reply Favorite

Date: June 4th, 2026 11:39 AM
Author: multi-colored box office



(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914742)



Reply Favorite

Date: June 4th, 2026 11:48 AM
Author: Thirsty judgmental location french chef

Everything else is made in China. China provoded tons of services that use our data. There's no reason to think AI will be different unless the government prohibits it. Companies will always eventually go with the lowest cost thing that does the job at an enterprise level.

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914775)



Reply Favorite

Date: June 4th, 2026 11:51 AM
Author: Histrionic swollen old irish cottage

its more comparable to telecom prohibitions than importing rubber trash cans

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914792)



Reply Favorite

Date: June 4th, 2026 11:55 AM
Author: Thirsty judgmental location french chef

Why are the telecom prohibitons necessary if corp wouldn't use them?

Because, like I said, they will use it unless it's prohibited.

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914797)



Reply Favorite

Date: June 4th, 2026 11:32 AM
Author: aggressive brass community account

the US government will just make it illegal to use chinese AI models

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914711)



Reply Favorite

Date: June 4th, 2026 11:35 AM
Author: diverse ticket booth kitty

They are all distilled counterfeit garbage anyway

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914723)



Reply Favorite

Date: June 4th, 2026 11:36 AM
Author: Histrionic swollen old irish cottage



(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914726)



Reply Favorite

Date: June 4th, 2026 11:37 AM
Author: aggressive brass community account

yeah but he's right that a lot of companies would otherwise use them anyway because they're an order of magnitude cheaper

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914729)



Reply Favorite

Date: June 4th, 2026 11:40 AM
Author: Histrionic swollen old irish cottage

enterprise foundation models dont charge per token except for frontier capabilities

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914745)



Reply Favorite

Date: June 4th, 2026 11:55 AM
Author: diverse ticket booth kitty

They can just use cheap US models. What jobs really need frontier reasoning?

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914800)



Reply Favorite

Date: June 4th, 2026 11:38 AM
Author: razzle meetinghouse

they can't and won't

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914736)



Reply Favorite

Date: June 4th, 2026 11:43 AM
Author: aggressive brass community account

ofc they can, the whole US economy is going to revolve around AI soon

they will be forced to even if they didn't want to

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914757)



Reply Favorite

Date: June 4th, 2026 11:44 AM
Author: Histrionic swollen old irish cottage



(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914763)



Reply Favorite

Date: June 4th, 2026 11:48 AM
Author: Thirsty judgmental location french chef

They can and won't.

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914777)



Reply Favorite

Date: June 4th, 2026 1:17 PM
Author: razzle meetinghouse

no, they can't. we are completely beholden to china wrt trade and this is an important enough arms race that the chinese government would buttfuck us if we tried.

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49915036)



Reply Favorite

Date: June 4th, 2026 8:04 PM
Author: Thirsty judgmental location french chef

This is all reasoning for the "won't" bucket. We certainly have the ability to.

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49916096)



Reply Favorite

Date: June 4th, 2026 11:55 AM
Author: Obsidian Marvelous Laser Beams University

has anyone tried coding with chinese models?

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914798)



Reply Favorite

Date: June 4th, 2026 11:56 AM
Author: diverse ticket booth kitty

Yeah it's not horrible but now that codex 5.5 is insane you can't really just go back

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914803)



Reply Favorite

Date: June 4th, 2026 12:00 PM
Author: Obsidian Marvelous Laser Beams University

is it really that much cheaper or better than claude code?

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914820)



Reply Favorite

Date: June 4th, 2026 12:02 PM
Author: diverse ticket booth kitty

Yeah it's significantly cheaper Open AI appears to be heavily subsidizing the cost. It's like twice as cheap for input tokens and 20-40% cheaper on output depending which gpt pricing tier you aare using. Also not just for tokens for the 100 dollar plan on codex vs claude you get way more usage. Claude depletes faster during peak hours too.

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914828)



Reply Favorite

Date: June 4th, 2026 5:46 PM
Author: Orange misanthropic church skinny woman

They nerfed codex usage limits massively on the 1st

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49915825)



Reply Favorite

Date: June 4th, 2026 12:05 PM
Author: Thirsty judgmental location french chef

How many times have people said "Have you ever tried using a Xhinese X?" about a products that are now exclusively produced in China?

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49914841)



Reply Favorite

Date: June 4th, 2026 5:53 PM
Author: Stimulating drab ape

I don't have coding needs (until I try to make a video game). But I've used deepseek for a bunch of basic, non-legal shit, and it generally sucked

The only good point I can think of is that it wouldn't bullshit when it was caught not following the prompt or not pulling data. ChatGPT and most American models dig their heels in and seem to intentionally fuck up requests once they've been caught

An example is that I was using AI for some dividend screening help. Despite explicit instructions to conduct research by calculating yields based on the prior day's closing price, it would pull 3 and then make up the rest (about 7 in total). When called on it, it would admit to the problem of not pulling data. I would explicitly tell it to pull the data for all tickers to generate the yield data. It would regenerate the data but refuse to pull the actual data for multiple tickers each time.

After four rounds of this, it finally suggested starting all over with a clean prompt tomorrow

If the Chinese models can get better quality-wise, they will fucking eat these lazy American LLMs for lunch. But that day is not yet here

(http://www.autoadmit.com/thread.php?thread_id=5871022&forum_id=2Vannesa#49915832)