Steven Yin

speedrunning @xai prev: macrodata refiner @openai, HFT @weHRTyou

blog:

[Steven Yin @stevenydc]

Trump card sales 📈

[Anthropic @AnthropicAI]

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.

Access to all other Claude models is not affected.

We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.

Read our full statement: https://t.co/bwn0sximKZ

[Steven Yin @stevenydc]

Bullish on ex Quants in AI. They have high agency (taking bit pay cuts and moving industry, often starting long distance relationship with partner) and they are “double major”

[TBPN @tbpn]

.@Collision is bullish on two types of people: high-agency individuals and double majors.

"There are two categories of people I would be super bullish on right now and I think will do incredibly well over the next 10-20 years. First, high-agency people. The people at Stripe who have been talking to customers and know exactly what we should do. It's the people who have that pep in their step and want to go make Stripe better. They are so much more empowered thanks to AI."

"The second is double majors. I think if you understand software and understand finance, or if you understand software and understand marketing, you now can go massively improve the entire marketing funnel for your company. Now, one person can do what would have taken 20 people dredging through all these systems."

"Charlie Munger talked about the importance of being multidisciplinary and multidisciplinary thinking. He thinks getting a functional understanding of many disciplines is not that hard. You can just go read the books now or you can talk to your AI about it. I think multidisciplinary thinkers are going to do incredibly well."

[Steven Yin @stevenydc]

@alz_zyd_ Most people should start asking questions instead.

[Steven Yin @stevenydc]

And then it’ll just be central planning

[François Chollet @fchollet]

One of the first signs of the emergence of superhuman AGI will be the emergence of a quantitative trading firm with impossible returns

[Steven Yin @stevenydc]

In trading, “everyone is a risk manager”

[Elon Musk @elonmusk]

@TheZvi Because everyone’s job is safety. It’s not some fake department with no power to assuage the concerns of outsiders.

Tesla has no safety team and is the safest car.

SpaceX has no safety team and has the safest rocket. Dragon is what NASA trusts most to fly astronauts.

[Steven Yin @stevenydc]

The self serve line is always 10x longer than the pre-served line at the same station at oAI. I never understood the premium for self serve TBH.

[Alexandra Barr @BarrAlexandra]

as a microcosm of culture, food at oai is always self serve and at anthropic it’s pre-served

it’s possible germs are spread more easily at oai (collective action problem) but i enjoy the autonomy of not having to ask permission for a second serving

[Steven Yin @stevenydc]

is neural link the final modality

[Steven Yin @stevenydc]

In the arc of human history, building microwave towers to beam prices around turned out to be a pretty epic side quest

[Noam Brown @polynoamial]

@_sholtodouglas @jekbradbury @GoogleDeepMind @andy_l_jones @OpenAI @kevin_wang3290 A word on comp: I know folks who become quants to make money but 5 years later ask what they're doing with their life. We're at a special time in history. In AI research, you can help positively guide the most important tech of our time *and* get paid well https://t.co/oOZAoQXsXO

[Steven Yin @stevenydc]

Ads is probably the only way to monetize consumer AI products long term: cost/intelligence is going to be competed down to ~0 and average user does not need that much intelligence.

[Steven Yin @stevenydc]

@iaindunning What’s this number going to be 5 years from now?

[Steven Yin @stevenydc]

Working on math at frontier labs feels like a status symbol thing. Similar to humanities departments at universities.

[Steven Yin @stevenydc]

Gradually, then suddenly

[Sebastien Bubeck @SebastienBubeck]

This problem fits in a broader context of understanding THE SHAPE OF LEARNING CURVES. The most basic property of such shapes is that hopefully ... they are decreasing! Specifically from the statistical perspective, assume that you add more data, can you prove that your test loss will be lower?

Surprisingly this is quite non-obvious and there are many counterexamples. This was discussed at length in the classic book [Devroye, Gyorfi, Lugosi, 1996] (which I remember reading voraciously 20 years ago but that's a different story!). More recently in a 2019 COLT Open Problem it was pointed out that some extremely basic versions of this question are still open, such as: if you estimate the (co)variance of an unknown Gaussian, is the risk monotone (ie adding more data helps you estimate this covariance better)??

@MarkSellke asked this question to GPT-5.2 and ... it solved it! And then Mark engaged in a back and forth with the model to keep generalizing the result (with no mathematical input from Mark except asking good questions) and it kept going ... eventually this became a nice paper, with results for both Gaussian and Gamma distributions for forward KL, and more general exponential families for reverse KL. You can read more about it here: https://t.co/XLETMtURcd .

[Steven Yin @stevenydc]

Yet is the key word!

[Ethan Mollick @emollick]

What a fascinating, frustrating paper. It makes it clear that GPT-5 is capable of new discoveries in challenging fields, but that this process currently requires guidance and expertise.

The frustrating part is there is no repeatable method identified for others to follow (yet).

[alz @alz_zyd_]

you think LLMs will revolutionize math education. But math has always been, in a sense, trivial. There have always been 13-year-olds who know real analysis. The binding constraint has always been willpower, that is, of parents to push their moderately talented kids unusually hard

[Sam Altman @sama]

A first preview of something we expect to see a lot more of soon:

[Steven Yin @stevenydc]

The Codex moment for science is here

[Kevin Weil 🇺🇸 @kevinweil]

💥 Today we say “hello world” from OpenAI for Science.

We’re releasing a paper showing 13 examples of GPT-5 accelerating scientific research across math, physics, biology, and materials science. In 4 of these examples, GPT-5 helped find proofs of previously unsolved problems.

[Steven Yin @stevenydc]

@hirschibar I actually don’t know

[Steven Yin @stevenydc]

kv cache is the new ads real estate