Jérémy Andréoletti 🔸

Looking back on 2025: what were our most popular short-form research posts? We published 36 Data Insights and 37 Gradient Updates newsletters. Here's what our website readers found most interesting: 🧵

New paper by my colleague @antmaier at the GPAI Policy Lab! I love how he synthesized and formalized many classical arguments, showing that i) current AI systems don't internalize their developers' goals, and ii) over-optimizing powerful AI models is dangerous.

What will AI look like by 2030 if current trends hold? Our new report zooms in on two things: (1) whether scaling continues (compute, data, power, capital), and (2) the capabilities this enables—especially for scientific R&D.

Great work led by @JyAndreoletti, on the Occurrence Birth-Death with lineage-specific speciation and extinction rates!

I'm very impressed with the Sentinel newsletter: by far the best aggregator of global news I've found Expert forecasters filter for the events that actually matter (not just noise), and forecast how likely this is to affect eg war, pandemics, frontier AI etc Highly recommended!

Happy to share our new preprint: `The rise, decline and fall of clades' with @JyAndreoletti, Daniele Silvestro and @HMorlon , where we study how the biodiversity of clades rise and fall, and the processes behind these patterns. 1/5 doi.org/10.1101/2025.0…

Most OECD countries fail to reach the UN’s target for aid to the poorest countries— In 1981, the major foreign aid donor countries, also known as the Development Assistance Committee, made a promise at the UN: to aim for at least 0.15% of their national income to assist the world's least developed countries — about 1 dollar out of 700. Over the years, they repeated this pledge. But by 2022, most countries failed to honor this promise. The chart shows the 20 OECD countries giving the most aid, as a percentage of their national income, to the world’s poorest nations. Only three countries met the target: Luxembourg, Sweden, and Norway. Some countries are so wealthy that even a rounding error in their budgets could mean the difference between life and death for people in the poorest parts of the world. (This Daily Data Insight was written by @S_VanTeutem.)

In honor of international women's day, let's take a moment to remember the most famous statistician in history. You've definitely heard of her, but you probably have no idea she was a statistician. It's Florence Nightingale.

Indoor air quality has been a problem for most of human history. From Viking longhouses to 19th-century London, the struggle to breathe clean air has shaped how we build, live, and even govern. This is the BATTLE FOR BETTER AIR. 🧵

Flu pandemics have appeared every 10 to 50 years for millennia. Vaccines have been available for 80 years, but reduce infection risks by just 40% on average. How can we make "universal" vaccines that confer lifelong immunity against any strain of flu: past, present, or future?

I've mentioned previously that one of the more dangerous carbon monoxide scenarios is flue pipe disconnect because it doesn't initiate internal equipment safety switches. If a disconnected appliance produces high CO, functioning CO detectors are the last line of defense

People sometimes ask me why I am so obsessed with lead poisoning - and sometimes I ask myself that too - but this paper sums it up well. - 5.5 million premature adult deaths from cardiovascular disease - US$1.4 trillion in global economic losses - we can solve it

Do you have an idea to tackle a critical bottleneck in science and technology that cannot be addressed through current institutional structures? Our Activation Partner, @Convergent_FROs, is seeking proposals that could unlock transformative progress in our opportunity spaces👇

"Announcing biosecurity.world" by Alix Pham (@alix_ph), Swan, Jérémy Andréoletti (@JyAndreoletti), and johantang🔸

This project was part of the Biosecurity Fundamentals: Pandemics Course by @BlueDotImpact, which I highly recommend for anyone interested in the topic! biosecurityfundamentals.com/pandemics 🥈It won runner-up in the 'Novel Research' category! (11/11)

- What alternative non-modelling interventions would be even more effective in achieving some of these objectives? Conclusion: Epidemics are inevitable, but their impact doesn’t have to be! (10/11)

[𝗡𝗲𝘄 𝗽𝗼𝘀𝘁!] 𝗖𝗿𝗶𝘁𝗶𝗰𝗮𝗹 𝗥𝗲𝘃𝗶𝗲𝘄 𝗼𝗳 𝗠𝗼𝗱𝗲𝗹𝗹𝗶𝗻𝗴 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗲𝘀 𝗔𝗴𝗮𝗶𝗻𝘀𝘁 𝗣𝗮𝗻𝗱𝗲𝗺𝗶𝗰𝘀 Can we outsmart pandemics? I did a deep dive into the world of pandemic modelling tools to find the interventions that might actually help 👇 (1/11)