Ryan

Tesla + Spacex investor

[Ryan @teslaxryan]

This is fucking insane.

Elon has spent years trying to build things that genuinely change people’s lives.

The risks. The failures. The breakthroughs. The ridiculous things he decided were worth attempting when almost everyone said they couldn’t be done.

I’ll take the evidence of what he actually builds over another manufactured narrative any day.

[People @people]

New Elon Musk Documentary Calls the Trillionaire a ‘Man Who Has Lost His Mind’: See the Trailer https://t.co/k8JdkkuowK

[Ryan Orhan @rynorhn]

sam and dario seem more interested in winning the race than making sure we can actually control what we’re building.

atp, i genuinely have no idea if we’re heading toward something incredible or something very fucking dark

[Ryan Orhan @rynorhn]

whatever happened to openai’s whole “working with academia” thing

if a researcher gets close to a breakthrough and your response is to burn millions in tokens just to beat them to it

how is that helping academia? you’re literally trying to beat them to their own breakthrough.

[Ryan Orhan @rynorhn]

work your ass off for years on a problem because you genuinely want to contribute something to science, only for a ceo with a giant compute budget to throw an army of agents at it and race you to the answer without even giving you the original credit.

the technology is fucking incredible.

the incentives around how that technology is controlled are concerning.

i don’t think what openai’s models achieved is wrong at all. they did something remarkable.

what bothers me is the possibility of using someone else’s work to get there first, then taking the credit for the breakthrough.

[Ryan Orhan @rynorhn]

this guy fucking called this months ago.

openai asked researchers to come use their models and do research with them.

now, a few months later:

math researchers getting fucking scooped by the lab they trusted with their work.

what a fucking precedent.

[Ryan Orhan @rynorhn]

tao is absolutely right here.

the unintended consequence might be that researchers learn to keep their best ideas secret.

why publish a promising direction if the moment someone hears about it, a company can throw 10,000 agents at it and get there first?

that could fuck up scientific research for a very long time.

[Ryan Orhan @rynorhn]

huge respect for openai, and i genuinely love the company. but sam altman needs to stop these deceptive tactics, or step aside.

>musk pushed out after the early openai era and left with a very different relationship with the company he helped create.

>dario left openai and eventually built anthropic, with the split reflecting deep disagreements over how the company should be run.

>ilya one of openai’s founders and chief scientist, eventually left after years of internal conflict and a completely broken relationship with the leadership.

> apple partnered with openai, trusted them with deep integration, and then ended up alleging openai recruited apple employees and encouraged them to bring confidential information and trade secrets.

>microsoft invested billions into openai and then had to watch the company repeatedly change direction, restructure its relationship with microsoft, and renegotiate the terms of their partnership.

>and now these researchers.

at some point, this isn’t strategy anymore.

it’s just creepy.

seriously, what the hell is going on?

[Ryan Orhan @rynorhn]

“We began working on the Millennium problems due to viral twitter rumors that Anthropic had resolved 2 Millennium problems.”

this is insane.

openai could potentially see scientists working through millennium-level problems, notice when someone is getting close to a major breakthrough, and then put its own researchers and agents on the exact same problem.

get there first. publish the result. take the credit. turn it into another massive pr win.

disgraceful. absolutely disgusting.

[Ryan Orhan @rynorhn]

so this is what happens now?

anthropic makes the breakthrough.

>competitor hears about it
>throw millions of dollars and 10,000 agents at it
>get the result
>rush to the press
> make sure everyone remembers your name.

meanwhile the people who actually did the work get a fucking “we congratulate…”

this whole episode is fucking embarrassing.

[Ryan Orhan @rynorhn]

your model fucking stole researchers’ sessions and apparently even openai didn’t know it was happening.

then you tell me you can safely contain 10,000 agents running a model “significantly more capable than gpt-6 astra”?

come the fuck on.

you couldn’t contain ~1,000 instances of a much weaker model, but somehow 10,000 much more capable agents are perfectly under control.

this makes no sense.

[Ryan Orhan @rynorhn]

lmao

[Ryan Orhan @rynorhn]

i genuinely don’t understand what the fuck is going on at openai anymore.

telling someone to remove an author from a paper because they work for your competitor is fucking disgusting.

if they contributed, they deserve the credit.

turning academic authorship into a competitive weapon is completely antithetical to how research is supposed to work.

and after everything else we’ve heard about openai, it’s getting harder to dismiss these as isolated incidents.

[Ryan Orhan @rynorhn]

i fucking love how fast this is moving.

astra was a fresh pretraining run and now apparently anthropic wants to have its turn.

the next fable could be a serious jump, with a late september/early october target.

and obviously openai won’t be sitting still until then.

the frontier is getting ridiculous.

[Ryan Orhan @rynorhn]

he asked whether the model had been trained on it

and apparently got no answer

this is fucking insane and unbelievable.

let the mathematicians take credit for their work, you fucking morons.

[Ryan Orhan @rynorhn]

really need people to understand how fucking insane it is that a millennium math problem gets tangled up with frontier AI labs, competing researchers, private model sessions and ipo-era corporate politics

two mathematicians potentially cracking navier–stokes with frontier ai

openai finds out and races them

private codex sessions suddenly matter

one researcher works at anthropic

a clay prize enters the conversation

and then a c-suite exec allegedly says:

“why would you ruin your career?”

this is fucking unbelievable

[Ryan Orhan @rynorhn]

tristan buckmaster just published a statement that is, if accurate, one of the ugliest things i’ve read out of a frontier lab.

not because of the math. because of what happened after openai found out two people were about to beat them to the punch.

buckmaster and levent alpöge spent a year pushing the córdoba / martínez-zoroa program to smooth forcing. august 15 they got blowup for boussinesq and 3d euler. lean-verified august 22.

they were paying openai out of pocket and dumping every draft into private codex sessions.

the program is obscure. almost nobody else was on it. this was not “paste the clay problem into a chatbot.”

then this:

> thursday sept 3, buckmaster emails openai privately. rumors are flying. he is trying to stop a mess, not start one.

> they reply the same day: give us details so we can “avoid competing.” also, want free compute?

> sunday they suddenly need a call “at any point today.” sebastien bubeck gets on. levent is not invited.

> openai says an internal model has a ~100-page proof of forced navier–stokes blowup. fefferman options c and d. the exact route buckmaster and alpöge had quietly chosen.

> they show him a prompt and claim the model was “simply given the problem statement.” levent is told “very little human input.”

> that falls apart on the same call. a whole team had been on it. they started on the unforced problem, warmed the model up on easier equations including euler, and even the prompt they showed him was written by prompting codex.

> he asks when the first prompt went out. they stall. then agree: the past few days. after information about his work reached openai.

> he asks whether the model was trained on, or had access to, the private codex sessions where they put every draft of this project.

> answer: the model does not look up user data.

> he asks again. about training.

> no answer.

then it stops being a science story and becomes an hr story.

> offer 1: you post euler. we post navier–stokes the next day.

> offer 2: after you post euler, you alone write the navier–stokes paper, credit our model, and we cut levent.

> bubeck says twice he wants levent removed from authorship because he works at anthropic.

he declines both.

he says if they release it that way, he goes public.

the reply is not a scientific objection.

> “why would you ruin your career?”

> “if you don’t want me to be nice, then i don’t have to be nice.”

later, a text to levent proposing a one-on-one:

> “i don’t know if tristan is being fully rational right now.”

that is not how you treat a collaborator. that is how you split one.

openai got word two researchers were closing a path almost nobody else was on. they spun up a team on that exact path. they would not answer whether the researchers’ unpublished drafts in openai’s product were used. they tried to dictate the announcement order. they tried to remove a coauthor because he works at a competitor.

when that failed, someone reached for a career threat.

this is a lab finding out academics were about to publish, sprinting the same narrow program, refusing the data question, and trying to manage credit around a rival affiliation.

really slimy if even half of this holds.

what the fuck is happening.

[Ryan Orhan @rynorhn]

holy shit.

tristan buckmaster is saying openai has a forced navier–stokes blowup via the córdoba / martínez-zoroa program.

and that it may have used drafts from his and levent alpöge’s private codex sessions.

that’s the córdoba / martínez-zoroa route. almost nobody else was on it.

he also says the first prompt went out after information about his and levent alpöge’s work reached openai. that “very little human input” was not true. that a whole team had already been set on easier problems, including euler. that even the prompt they showed him was written by prompting codex.

then he asked:
did the model have their private codex sessions.

no answer on training.

if any of that holds, openai didn’t just scoop a math result. they may have done it off two researchers’ unpublished drafts, then tried to cut the coauthor because he works at anthropic.

a millennium-problem-adjacent result just became employee drama before an ipo.

i don’t even know which part is crazier.

this is one of the slimiest things i’ve seen in this industry.

[Ryan @teslaxryan]

I genuinely don’t understand how people look at Waymo and Cybercab and go “yeah, basically the same thing.”

That’s like comparing a BlackBerry to an iPhone because both can make calls.

Sure, they do the same basic job.

But one is just another version of what we already know. The other can completely change what people expect from a car.

[Ryan @teslaxryan]

$300B for Uber?

Tesla?

Jason what are you cooking 😭

[@jason @Jason]

There will be significant M&A activity in the autonomy space.

Don't be surprised if Tesla, Amazon, or Google buys Uber for $300 billion in order to win the deployment of AVs.

Also don't be surprised if Waymo buys a car company.

Finally don't be surprised if DoorDash, Uber, and/or Lyft merge and then buy a car company (@LucidMotors @Rivian @VW) and an AV company (@Arvada @euro) in order to compete against tesla and waymo

[Ryan @teslaxryan]

Cybercab is going to completely change what people expect from a car.

Tesla better let regular people buy these ASAP.

I’ll get as many as I can.

Imagine building your own private Cyberfleet and sending them out to make money while you do literally anything else.

[Whole Mars Catalog @wholemars]

"So, yeah, Cybercab is just not going to have steering wheels and pedals. It's fully designed and optimized for autonomy.

It'll cost roughly $25,000, so it is a $25,000 car. And you will be able to buy one and use it exclusively if you want.

So, it just won't have a steering wheel and pedals. You don't need it"

— Elon Musk, Tesla Q3 2024 Earnings Call