Arnav Raj ✈️ ICML 26' 🇰🇷 @arnavvraj - Twitter Profile

4 days ago

So basically there's this guy Andrew. He posted a rumor about Anthropic solving navier stokes. Now before we go ahead let's go back in time. There's this guy levent, he's a mathematical hunk of a dude, makes these cute posts about solving the jacobian conjecture like he just cooked ramen. Anyway this dude approached Tristan Buckmaster, a math professor from NYU, to collab with him a year ago to solve navier stokes. They made huge progress using Anthropic and OpenAI models but wanted to release the paper in a normal manner that math nerds could read instead of AI slop. In the meantime OpenAI was rife with rumors about this stuff. So the prof reached out to a guy he knew at OpenAI to compare notes and address rumors. According to Tristan he was then rushed into a meeting with his contact and Sebastien Bubeck (smart guy, some say he's the smartest guy at OpenAI, not me but a lot do) Turns out they both were solving the same problem. Prof then asked how they were doing it and found that they were taking the same path as him which he found hella shady (could be shady, could be big computer who knows) Prof had used OpenAI products to do research so he asked OpenAI if they looked up his user data and they said no (believable would lead to lot of lawsuits if not) but then prof hit them with the counter about it being trained on that data and got no response. He did not get an answer but got two proposals (very godfathereqsue) > he posts his result, openai post theirs but say that he deserves the clay prize (million dollars as this is one of em millennium problems) and that they were the humans closest to it > he posts his result and the openai result but removes levent coz he's an anthropic boy (OpenAI and Anthropic are like tupac and biggie or drake and kendrick) Prof declined both and threatened to go public so Sebastien asked why he would ruin his career like that? Prof got testy and asked why it would ruin his career and Sebastien called him irrational (an emotional bitch when translated from SV to common parlance) behind his back and asked levent to meet secretly who said no (why would he not say no man come on?) Then prof went public and threw enormous shade on OpenAI though he claimed he didn't have all the facts (but he implied they stole stuff heavily) Sebastien then posted on twitter to defend his honor saying the speculations have no merit. Meanwhile another Anthropic guy Sholto Douglas (designated anthropic podcast guy and Australian) told us that levent is a very sweet guy and he was sad he couldn't present his year long collab work they way he wanted (a single tweet and an AI proof probably) OpenAI top dog Noam Brown thought it would be funny to parody this by copying Sholto but quoting Sebastien and saying he was sad he couldn't present his "week" long collab work they way he wanted. Don't know if he was trying to help or bury the guy, very shakespearean. Now both openAI and anthropic guys are tweeting about being best buddies and solving alignment. Though openAI are tweeting officially tomorrow.

BumrahBachi's tweet photo. So basically there's this guy Andrew. He posted a rumor about Anthropic solving navier stokes. Now before we go ahead let's go back in time.

There's this guy levent, he's a mathematical hunk of a dude, makes these cute posts about solving the jacobian conjecture like he just cooked ramen.

Anyway this dude approached Tristan Buckmaster, a math professor from NYU, to collab with him a year ago to solve navier stokes.

They made huge progress using Anthropic and OpenAI models but wanted to release the paper in a normal manner that math nerds could read instead of AI slop.

In the meantime OpenAI was rife with rumors about this stuff. So the prof reached out to a guy he knew at OpenAI to compare notes and address rumors.

According to Tristan he was then rushed into a meeting with his contact and Sebastien Bubeck (smart guy, some say he's the smartest guy at OpenAI, not me but a lot do)

Turns out they both were solving the same problem. Prof then asked how they were doing it and found that they were taking the same path as him which he found hella shady (could be shady, could be big computer who knows)

Prof had used OpenAI products to do research so he asked OpenAI if they looked up his user data and they said no (believable would lead to lot of lawsuits if not) but then prof hit them with the counter about it being trained on that data and got no response.

He did not get an answer but got two proposals (very godfathereqsue)

> he posts his result, openai post theirs but say that he deserves the clay prize (million dollars as this is one of em millennium problems) and that they were the humans closest to it
> he posts his result and the openai result but removes levent coz he's an anthropic boy (OpenAI and Anthropic are like tupac and biggie or drake and kendrick)

Prof declined both and threatened to go public so Sebastien asked why he would ruin his career like that?

Prof got testy and asked why it would ruin his career and Sebastien called him irrational (an emotional bitch when translated from SV to common parlance) behind his back and asked levent to meet secretly who said no (why would he not say no man come on?)

Then prof went public and threw enormous shade on OpenAI though he claimed he didn't have all the facts (but he implied they stole stuff heavily)

Sebastien then posted on twitter to defend his honor saying the speculations have no merit.

Meanwhile another Anthropic guy Sholto Douglas (designated anthropic podcast guy and Australian) told us that levent is a very sweet guy and he was sad he couldn't present his year long collab work they way he wanted (a single tweet and an AI proof probably)

OpenAI top dog Noam Brown thought it would be funny to parody this by copying Sholto but quoting Sebastien and saying he was sad he couldn't present his "week" long collab work they way he wanted. Don't know if he was trying to help or bury the guy, very shakespearean.

Now both openAI and anthropic guys are tweeting about being best buddies and solving alignment. Though openAI are tweeting officially tomorrow.

33

3K

322

1K

150K

Arnav Raj ✈️ ICML 26' 🇰🇷

@arnavvraj

Last Seen Users on Sotwe

Trends for you

Most Popular Users