Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code. But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along. So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions. TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.

For my friends who are still using UV and might be a little weary about recent compromises to PyPi packages, stick this in your pyproject.toml. You can let all of those pip users find and report the compromises...

I am searching for postdocs interested in developing a deeper scientific understanding of any aspect of AI and using this understanding to improve AI systems. If interested please email me materials by Jan 10th 2026: a CV, 3 letters of reference, and a statement of research interests (can be short, just indicating what you would like to work on). The positions are generously funded by grants from @schmidtsciences and @SimonsFdn. You will have the opportunity to collaborate with many labs across @StanfordHAI and also the Simons Collaboration on the Physics of Learning and Neural Computation (physicsoflearning.org/).

This is a fantastic review. We are entering an era of discovering rare cosmic phenomena. cambridge.org/core/journals/…

Today, we’re showing off our favorite pictures of culpeo foxes we’ve spotted around Cerro Pachón, home of NSF–DOE Rubin Observatory. 🦊 📸 💡Fun fact: While these little guys look like foxes (and are called foxes), they’re actually more closely related to wolves and jackals!

🚀We’re looking for 2026 interns at @PolymathicAI (NYC)! Want to work on scientific foundation models ML for physics, biology, astronomy, & more? Want to contribute to frontier research with a brilliant, fun, and friendly team? Please sign up on our interest form 👉 forms.gle/hxXPCv3qssEkEQJT7 #Hiring #Internship #MachineLearning

SPICE -- modelling synthetic spectra of stars with non-homogeneous surfaces. arxiv.org/abs/2511.10998

Susumu Kitagawa, Richard Robson and Omar M. Yaghi have been awarded the 2025 #NobelPrize in Chemistry for the development of a new type of molecular architecture. In 1989, chemistry laureate Richard Robson tested utilising the inherent properties of atoms in a new way. He combined positively charged copper ions with a four-armed molecule; this had a chemical group that was attracted to copper ions at the end of each arm. When they were combined, they bonded to form a well-ordered, spacious crystal. It was like a diamond filled with innumerable cavities.

Thanks to everyone who came to our Australia Telescope Users Committee open meeting today!

Registrations are now open for our next Australia Telescope User Committee meeting (ATUC). ATUC is a chance to discuss issues that impact users of the ATNF. 16 September, online and in-person in Sydney. Register: forms.office.com/r/hdP8Qspix…

⚠️ Tsunami Warning: Japan’s East Coast A powerful magnitude 8.7 earthquake struck off the Russian coast, north of Hokkaido, this morning. Tsunami waves are now moving south across the Pacific and may impact Japan’s eastern coastline. If you’re in the affected regions, stay away from the sea. Follow official evacuation advice for your safety.

BREAKING: The state of Hawaii is now under a Tsunami Warning. Here's the latest:

TSUNAMI WARNING 3: See tsunami.gov for alert areas.  M8.7 080mi SE Petropavlovsk, Kamchatka 1625PDT Jul 29:

Air travel fuels ~$400B of U.S. GDP, but space weather like solar flares & winds can disrupt navigation, comms, & passenger safety. @NWSSWPC tracks the Sun’s activity and provides warnings and predictions in real-time, helping the United States’ 700,000 pilots make the most informed decisions in the air. Learn more at swpc.noaa.gov/communities/av… #SpaceWeather #Aviation

The Academy’s new 10-year plan for astronomy is being launched today in Adelaide at the annual meeting of Australia’s astronomers. Read more: science.org.au/news-and-even…

The @AstroSocAus honours six astronomers at its Annual Scientific Meeting in Adelaide this week, including our Dr Joshua Preston Pritchard - winner of the Emerging Leaders in Astronomy Software Development Prize. Read about all the winners: scienceinpublic.com.au/media…

Can AI help solve Millennium Math problem in Fluid dynamics? We take an important step in this direction in our recent paper with physics-informed learning: High-Precision PINNs in Unbounded Domains: Application to Singularity Formulation in PDEs We develop a modular and a robust framework for training Physics-Informed Neural Networks (PINNs) to high precision on unbounded domains. This is a key step toward using AI in rigorous PDE analysis and computer-assisted proofs of singularities. Highlights: 1. Enforcing far-field asymptotics and non-degeneracy via hard constraints Leveraging self-scaled BFGS optimizers for stable convergence. 2. Achieving 4-digit improvement over state-of-the-art for 2D Boussinesq with fewer training steps. 3. Applicable to both smooth and non-smooth self-similar blowup profiles We believe this is a significant step toward integrating machine learning into the toolbox for tackling open problems in fluid dynamics and singularity formation (e.g. Navier-Stokes). arxiv.org/abs/2506.19243 @Caltech

"Training deep learning models on petabyte-scale Earth observation (EO) data requires separating compute resources from data storage. However, standard PyTorch data loaders cannot keep modern GPUs utilized when streaming GeoTIFF files directly from cloud storage. In this work, we benchmark GeoTIFF loading throughput from both cloud object storage and local SSD, systematically testing different loader configurations and data parameters. We focus on tile-aligned reads and worker thread pools, using Bayesian optimization to find optimal settings for each storage type. Our optimized configurations increase remote data loading throughput by 20× and local throughput by 4× compared to default settings. On three public EO benchmarks, models trained with optimized remote loading achieve the same accuracy as local training within identical time budgets. We improve validation IoU by 6–15% and maintain 85–95% GPU utilization versus 0–30% with standard configurations. Code is publicly available at github.com/microsoft/pytorch…"