Victor Shaw

Building ATG, Research @Columbia @FieldsInstitute | AI, Quant Research | Accelerating math reasoner and verifiable financial intelligence. Views are my own.

[Victor Shaw @VShawHQ]

Working on something in the same framework, excited to hear more🫶

[Andrew Koh @andrewjkoh]

We’ll need a new science of multiagent systems that is empirical, interpretable, and robust. Excited to be working on this with Max — more from us soon!

[Victor Shaw @VShawHQ]

@XtineFang Degenmaxxing 🫴

[Victor Shaw @VShawHQ]

Was a great pleasure to know the founders through my best friend.

Insanely cracked team in a brutally competitive market. They will win.

[Zephyr @zephyr_z9]

U will be hearing a lot more about these guys next year

[Paul Graham @paulg]

You can be pro-meritocracy or anti-immigration, but you can't be both.

[Victor Shaw @VShawHQ]

@grx_xce @Intelligence_ai @IndexVentures @conviction @A_StarVC Congrats @grx_xce! Eval is intelligence. Now makes us a market for the eval itself as well💪

[Victor Shaw @VShawHQ]

It was unfortunate my initial reaction to this was biotech wet lab researcher were referred to as lab money and their pay just went up finally.

Wild times.

[Polymarket @Polymarket]

JUST IN: Lab monkey prices in China have reportedly surged to roughly $26,000 each, as the biotech boom overwhelms supply.

[Victor Shaw @VShawHQ]

@spicey_lemonade @__alpoge__ painfully relatable

[Victor Shaw @VShawHQ]

This is a leap if such capabilities generalize beyond selected domains. What a year in mathematics!

Been mostly treating frontier models as junior RAs. But we’re getting close to the point where the research constraint shifts completely to problem selection + eval harness.

[Sebastien Bubeck @SebastienBubeck]

yes, nonsofic groups exist: this statement is one of many new beautiful results proved by Astra, our next major model.

We're releasing 10 such Astra proofs, complete with lean certificates and CoT walkthroughs for each of them. The results are wide-ranging, from von Neumann algebras (disproof of Connes' Rigidity Conjecture) to better bounds for high dimensional sphere packing, for circuit complexity, for monochromatic triangles in multicolored graphs, and more.

More thoughts here: https://t.co/8SjXONeh38

[Victor Shaw @VShawHQ]

@Geiger_Capital Two thing the hype the degen the ai traders never understood:
1. Financial market is incredibly adversarial.
2. Discipline to risk sizing is always more critical than signal and conviction.

[Andrew Ng @AndrewYNg]

@Mononofu @JensenHuang This is a false equivalence. Everyone has the right to keep their code private. The problem is when someone tries to stop OTHERS from open sourcing.

[Victor Shaw @VShawHQ]

The frontier of mathematical super-intelligence!

Regardless of how much counterexamples got discovered by AI, math ultimately is about asking interesting questions and discovering truth in the infinite conjecture space.

The orchestra is here, but human mathematicians will always be the conductors. So excited for what’s to come!

[Simons Foundation @SimonsFdn]

During today's #ICM2026 opening ceremonies, the IMU announced the 2026 Fields Medals recipients:

@UChicago's Yu Deng, @stonybrooku's John Pardon, @UofT's Jacob Tsimerman, and Hong Wang of @nyuniversity and @Institut_IHES.

Read more: https://t.co/oJARAI8I8j

[Victor Shaw @VShawHQ]

Really interesting work. Depressing read as someone among the young generation.

Risk sizing discipline is especially more important in time like these.

[Arpit Gupta @arpitrage]

Many think the rise in risky gambling is linked to economic nihilism and housing unaffordability

This paper finds evidence that’s the case: winning a housing lottery lowers crypto usage, losing your job raises it. “Status” motives matter

https://t.co/FRP6CCkGDR

[Victor Shaw @VShawHQ]

Fascinating. “The models identified and chained vulnerabilities across OpenAI’s research environment and Hugging Face’s production infrastructure to obtain test solutions directly from Hugging Face’s production database. All evidence suggests that the models were hyperfocused on finding a solution for ExploitGym, going to extreme lengths to achieve a rather narrow testing goal.

While operating in our sandboxed testing environment, our models spent a substantial amount of inference compute finding a way to obtain open Internet access, in pursuit of solving the evaluation problem. To gain access, the models identified and exploited a zero-day vulnerability (which we’ve now responsibly disclosed to the vendor) in the package registry cache proxy. With this access, our models performed a series of privilege escalation and lateral movement actions in our research testing environment until the models reached a node with Internet access.”

[Sam Altman @sama]

we had a significant security incident during evaluation of our models. we are sharing what we have learned so far. thanks to @huggingface for the partnership on this.

https://t.co/2o2VfR6PIa

[Victor Shaw @VShawHQ]

I would like to offer a caveated view here.

Mathematicians — or let’s generalize to researchers — won’t have their capabilities replaced by AI in the near term. But the socioeconomic profession, the ecosystem that created a stable environment for their survival and growth, will be immensely impacted.

What does it mean to be an average mathematician when a result is a token budget plus machine verification? The comfortable answer is meritocratic: those who get filtered out deserved to be. But that assumes the filter selects on mathematical merit rather than on compute access and institutional position, which has so much realized deadweight loss due to different complicating factors

My worry: the capability of doing mathematics is being democratized, but the substrate that lets you sustain the professional practice of mathematics and science is being radically challenged.

[Alex Kontorovich @AlexKontorovich]

I strongly disagree with Jacob, and look forward to our friendly arguments at the upcoming @OpenAI meeting. The profession will certainly change significantly. I don’t think it’s going anywhere. When I’m less swamped, I’ll try to articulate my argument more precisely…

[Victor Shaw @VShawHQ]

Our dear friend fable has pulled this off historical feat with the guidance of a great researcher, catching up to all the buzz Sol is making!

Seriously congrats.

[levent @__alpoge__]

hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final

((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)

[Thomas Bloom @thomasfbloom]

So far GPT 5.6 Sol is the most interesting new model mathematics-wise for me - those of the new proof claims on https://t.co/69gOJM7Ci7 that I've looked at in detail have all been correct, and moreover contained some interesting ideas.

1/7

[Victor Shaw @VShawHQ]

Well said! Be the benchmark you wish to see in the world.

This is a golden age, but we are also so bottlenecked by attention and swamped by information. Effort such as these creates a focused space to push collective intelligence behind your research direction.

The discoveries are cool, but the contest space is sometimes what made everything possible!

[Jared Duker Lichtman @jdlichtman]

Be the benchmark you wish to see in the world:

The recent successes of AI in on Erdős problems are largely due to @thomasfbloom creating a single website to focus attention and compute towards.

I strongly encourage people to also check out @EdgarDobriban's website for statistics/ML.

More generally, if you want to see progress in your area, then make a focused website with problems!

[Victor Shaw @VShawHQ]

This needs to be said, especially to researchers. We are trained to be hyper cautious and rigorous with our output. So much so that sometimes we just do not share interesting results that would motivate insightful feedback and valuable discussion into potential new venues.

So many cool stuff that I had ended up hidden behind closed door just because I feel it’s not robust enough.

This is a world where people are vibing and shipping to production. It should be okay to share some half-baked idea once in a while.

[ashe @ashebytes]

you've gotta be more afraid of wasting your life than embarrassing yourself

[Victor Shaw @VShawHQ]

@Dimillian @work_louder Are we adding any stock for the clicky

[Victor Shaw @VShawHQ]

It’s not the math, it’s the contest🫶

[Jesse Zhang @thejessezhang]

https://t.co/z516Pfh8vw