
@Google DeepMind. On leave, Canada CIFAR AI Chair and Former Research Director, @VectorInst. Professor, @UofT (Statistics/CS). Views are my own.
London
Joined June 2009
- Tweets28.3K
- Following2.1K
- Followers69.3K
- Likes22.8K
Pinned Tweet
Proud to have contributed to Gemini 4 Argon’s much improved math capabilities. It’s been an intense few months!
Replying to @Google
Gemini 4 Argon is our next era of frontier intelligence.
It shows significant improvements across benchmarks, setting a new state of the art for real-world long-horizon software engineering tasks.
Dan Roy retweeted
The question of information complexity of VC classes kept me busy for a year. In 2021, we made some progress: arxiv.org/abs/2111.05275
🎉 Now, it’s great to see Steve Hanneke and Juexiao Wang close the gap with a positive result in their FOCS ’26 paper: arxiv.org/abs/2610.10600
Dan Roy retweeted
This bug was first found and fixed by Gemini Flash 3.7, and most experiments in the paper were done by Flash 3.8. Gemini 4 Argon wrote the first draft of the paper and figures, which I aggressively nitpicked until the content was solid but the writing was not. Then I fed Argon a handful of my first-author papers from the last few years and had it do a rewrite in my style (which worked astonishingly well), and then I manually did a rewrite (the hands + keyboard variety) after that. I really liked this workflow! If you like my writing style feel free to throw some of my old papers in your context window and do the same.
How many algorithms (that were discarded) actually work? If you have your algorithmic loves that didn’t pan out, time to go back and double check?
Replying to @jon_barron
The first bug is "Softmax Collapse", first identified (but not directly fixed) by Prieto et al. 2025. When one logit is much larger than the others, its gradient silently rounds to zero (see figure), but the others *don't*, causing gradients to not sum to zero. Here's the fix:
Beautiful new result in statistical learning theory and probability: arxiv.org/abs/2610.24336
Dan Roy retweeted
If you'd like to come work with me and my amazing colleagues as a full-time researcher @MSFTResearch @MSRNE, please apply here:
apply.careers.microsoft.com/…
apply.careers.microsoft.com/…
Dan Roy retweeted
Today @GoogleResearch is launching Deep Review — an agentic system designed to stress-test manuscripts, design docs, and complex proposals before submission or publication. After pilot partnerships analyzing over 15,000 research papers, we’re making Deep Review globally accessible to help researchers, engineers, and professionals bring structured critique to their daily workflows.
Common sense recommendations.
Dan Roy retweeted
Over 400 academics in 30+ countries signed our open letter calling for immediate changes in AI development and deployment practices to ensure that cybersecurity incidents do not continue to occur, as future incidents will be far more damaging to our critical infrastructure.
Dan Roy retweeted
If you're a PhD student interested in interning with me or one of my amazing colleagues at Microsoft Research New England (@MSRNE, @MSFTResearch) this summer, please apply here apply.careers.microsoft.com/…
Dan Roy retweeted
Most important, we want to get this model into as many scientists’ hands as we can.
Today, we’re announcing that 10,000 scientists will get the highest usage tier access to Gemini 4 Argon in Google @antigravity free of charge for 6 months.
Apply here: forms.gle/WvmBbPvRbpKG1BPn9
Dan Roy retweeted
Great results from an exceptional model.
A great example of the positive impact of AI. This is why we at @GoogleDeepMind make advancing science a core focus!
Google Deepmind (FNV3) CRUSHED the forecast for #Isaias up to 11 days in advance. It picked up some appreciable signal on September 23, locked in on the exact landfall location4 days before it happened, and nailed the peak intensity consistently 3 days before it happened. A new age of forecasting is upon us.
First video: Track density using 1000 member ensemble
Second Video: Individual tracks and mean using 50 member ensemble
Dan Roy retweeted
Zurich airport, my backpack gets checked, the security agent takes out the copy of the little book of deep learning that I always have with me, goes through it rapidly, puts it back, and tells me "that's okay, you can go."
Huge news here in the UK.
NON COMPETES BANNED
Today, the PM has announced that his Government will legislate to functionally ban non-compete agreements for startups and scaleups in the UK or as he called it - the Bosman ruling for ruling for the innovation sector' 😂
At the @StartupCltn been working on this for months and we think it’s a huge potential unlock for our tech ecosystem.
Here’s what you need to know…
- Measures to functionally ban non-competes for startups and scaleups in the UK are going to be set out.
- There will also be consultation on a wider set of restrictive practices including gardening leave.
- More detail on these plans at the budget in a few weeks (Wednesday 28th October) including a timetable for consultation with a view to legislation in the coming months.
We think this is a huge move and so do lots of friends including so many of the great founders that signed the letter with Inherent just the other week.
Why? Because talent matters and mobile talent means startups can build. The Valley was built on free-flowing talent. In the UK, we’re now seeing our own versions of this story with companies being built from alumni of both homegrown firms like Revolut and beyond. But we need that to move faster. Hiring and building faster means British startups and scaleups can grow faster. Barriers to that slow our ecosystem down.
We won’t pretend it will be perfect, and people (including some of our friends but also lots of fairly vested interests too) will have a right to disagree. But we think that startups will thrive and win in a faster paced Britain.
So we are delighted with the progress today - as part of a strong speech by the PM outlining his ambition for building the world’s leading innovation nation including more on tax, growing companies outside London and more. But this is the start, not the end of the process. The announcement launches a consultation. It doesn't mean the issue is settled. How far reform goes, and what it looks like on the ground, is still to be decided.
The Prime Minister today has done something that can materially impact the trajectory of the UK tech ecosystem and with it our economy - but only if we make sure it’s delivered. So, the real work starts now.
Dan Roy retweeted
I have always believed that advancing mathematics is about empowering mathematicians.
From our earliest days at @GoogleDeepMind, our focus has been on deep collaboration with the mathematical community, pioneering new tools alongside one another to push the boundaries of what’s possible.
We’re releasing a broad range of new mathematical results produced by an internal frontier model.
We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results.
github.com/openai/math






















