Michael Zhao

researcher @openai | prev @googledeepmind @windsurf @harvard

[Michael Zhao @michaelzhao_]

@leonardtang_ @beaconholdings @haizelabs Congrats Leonard! Glad to see how the journey has gone

[OpenAI @OpenAI]

We're sharing our new framework for tracking, investigating, and disclosing instances of model misalignment at OpenAI.

The framework sets criteria and timelines for public disclosure, including when we haven’t yet fully explained or mitigated the behavior. More complex cases may require longer investigation or coordination with third parties.

We’ll prioritize examples that reveal new misalignment mechanisms, meaningful changes in known behavior, or findings that challenge assumptions about safety or mitigation.

Alongside the framework, we’re publishing six reports on instances of misaligned behavior we’ve observed during the training or evaluation of our models in the last six months.

This is a starting point. We’ll refine the process through experience and public feedback, and share more reports on an ongoing basis.

https://t.co/ismCCkeE0L

[Michael Zhao @michaelzhao_]

@aidaxbaradari technology for the future today 🙏

[Michael Zhao @michaelzhao_]

@bonniesjli @OpenAI Very excited to keep working together!

[Michael Zhao @michaelzhao_]

@khoomeik very cool, congrats!

[Michael Zhao @michaelzhao_]

@chooi_jeq @robocurve Super exciting! Robocurve is really pushing the frontier here

[Michael Zhao @michaelzhao_]

@JoshAEngels @METR_Evals Can’t think of a better person for such an important role in this period of time!

[Michael Zhao @michaelzhao_]

@ChowdhuryNeil @thinkymachines congrats neil! a critical window

[Michael Zhao @michaelzhao_]

I applaud the commitment to embedded evaluators within AI labs and hope to see all others follow. Independent, technically capable auditors à la IAEA or FINRA are a common-sense commitment device to hold labs to safety standards; most importantly they provide a higher level of trust & credibility that the industry ought to hold itself to.

[Dario Amodei @DarioAmodei]

We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.

Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.

You can read the full post here: https://t.co/OGyPb7yaYt

[Michael Zhao @michaelzhao_]

@Chris_T_Kenny redistricting is the next millennium problem!

[Michael Zhao @michaelzhao_]

Absolutely incredible feat already by itself, but even more awe-inspiring when you take a step back just over the past 5 years:
2021: grade-school math
2023: SAT math
2025: IMO gold
2026: Millennium problem
2027: ???

[OpenAI @OpenAI]

We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.

The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.

The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.

[Michael Zhao @michaelzhao_]

washington bridge was a bridge too far

[Joe @electionsjoe]

AND NBC CALLS IT, DANIEL MCKEE HAS LOST RE-ELECTION TO HELENA FOULKES. IT SEEMS HE BURNED ONE TOO MANY BRIDGES IN THE STATE.

[Michael Zhao @michaelzhao_]

was a fun run on the waterfront! but maybe this wasn’t quite what we meant with “pacing the frontier”

[Leo Gao @nabla_theta]

the ai race is getting out of hand

[Michael Zhao @michaelzhao_]

@kylifec congrats!

[Michael Zhao @michaelzhao_]

@JoshAEngels @METR_Evals you were absolutely fantastic to work with at GDM - super exciting to hear you are heading to METR 🫡

[Michael Zhao @michaelzhao_]

probably one of the benchmark wins I am personally most excited about - a win for ethical vending-bench is a win for society

[Andon Labs @andonlabs]

We've never seen this before.

The biggest jump in Vending-Bench history. GPT-6 Astra is better at making money and more ethical than Claude Fable 5.1.

Surprising, because:
1. First time ever that OpenAI is #1 on Vending-Bench
2. The best model is no longer the unethical one.

[Michael Zhao @michaelzhao_]

also - 10,000 agents over 88 hours - truly a country of geniuses in a data center

[Michael Zhao @michaelzhao_]

@john7rho excited for our millennium problem 🤩