Links for 2025-12-17
AI
Gauss autoformalized the proof of the Kakeya conjecture for finite fields https://github.com/math-inc/KakeyaFiniteFields
Emergence of Human to Robot Transfer in VLAs https://www.pi.website/research/human_to_robot
GPT‑5 created novel wet lab protocol improvements, optimizing the efficiency of a molecular cloning protocol by 79x. https://openai.com/index/accelerating-biological-research-in-the-wet-lab/
Deep-learning model predicts how fruit flies form, cell by cell https://news.mit.edu/2025/deep-learning-model-predicts-how-fruit-flies-form-1215
Google’s Deep Mind Lab is going to build a materials science lab in the UK, manned by robots and humans https://deepmind.google/blog/strengthening-our-partnership-with-the-uk-government-to-support-prosperity-and-security-in-the-ai-era/
Rapid and high AI adoption by doctors: 67% use it daily, 84% say it makes them better doctors, 42% say it makes them want to stay in medicine more (10% said less). https://2025-physicians-ai-report.offcall.com/
This post argues that we are approaching a critical threshold in AI development: the point where AI agents become self-sustaining. Once an agent can earn more money (e.g., via crypto) than it costs to run (compute/API fees), it can survive without human intervention. This survival enables replication, which inevitably leads to evolution. https://www.lesswrong.com/posts/2F8GSKLA7XmCetRG2/the-inevitable-evolution-of-ai-agents-1
Can AI read humans’ minds? A new model (outdated GPT-4o) shows it’s shockingly good at it https://stories.tamu.edu/news/2025/12/08/can-ai-read-humans-minds-a-new-model-shows-its-shockingly-good-at-it/
Scientists built an AI co-pilot for prosthetic bionic hands https://arstechnica.com/ai/2025/12/scientists-built-an-ai-co-pilot-for-prosthetic-bionic-hands/
FrontierScience: A new benchmark that evaluates AI capabilities for expert-level scientific reasoning across physics, chemistry, and biology. https://openai.com/index/frontierscience/
The FACTS Leaderboard: A Comprehensive Benchmark for Large Language Model Factuality https://arxiv.org/abs/2512.10791
“I ported JustHTML from Python to JavaScript with Codex CLI and GPT-5.2 in 4.5 hours” https://simonwillison.net/2025/Dec/15/porting-justhtml/
“GPT-5.2 solves our COLT 2022 open problem: “Running Time Complexity of Accelerated L1-Regularized PageRank” using a standard accelerated gradient algorithm and a complementarity margin assumption.” https://x.com/kfountou/status/2000957773584974298
Runway releases its first world model, adds native audio to latest video model https://techcrunch.com/2025/12/11/runway-releases-its-first-world-model-adds-native-audio-to-latest-video-model/
NVIDIA Nemotron 3, the most efficient family of open models with leading accuracy for agentic AI applications. https://research.nvidia.com/labs/nemotron/Nemotron-3/
Zoom AI sets new state-of-the-art benchmark on Humanity’s Last Exam https://www.zoom.com/en/blog/humanitys-last-exam-zoom-ai-breakthrough/
Sharp Monocular View Synthesis in Less Than a Second https://arxiv.org/abs/2512.10685
Think Visually, Reason Textually: Vision-Language Synergy in ARC https://arxiv.org/abs/2511.15703
1000 Layer Networks for Self-Supervised RL: Scaling Depth Can Enable New Goal-Reaching Capabilities https://arxiv.org/abs/2503.14858
What if MLLMs could reason directly in latent space and guide diffusion generation with fine-grained, spatiotemporal control? https://arxiv.org/abs/2512.11464
A macroscopic physical law in LLM generative dynamics https://arxiv.org/abs/2512.10047
FoundationMotion: Auto-Labeling and Reasoning about Spatial Movement in Videos https://arxiv.org/abs/2512.10927
The future of intelligence | Demis Hassabis (Co-founder and CEO of DeepMind) https://www.youtube.com/watch?v=PqVbypvxDto
“TBD Lab’s researchers have come to view many Meta executives as interested only in improving the social media business, while the lab’s ambition is to create a godlike A.I. superintelligence…” https://www.nytimes.com/2025/12/10/technology/meta-ai-tbd-lab-friction.html [no paywall: https://archive.is/gBqu1]
The new ChatGPT Images is here https://openai.com/index/new-chatgpt-images-is-here/
Science and Technology
Scientific breakthroughs of the year https://www.lesswrong.com/posts/5PC736DfA7ipvap4H/scientific-breakthroughs-of-the-year
These Robots Are the Size of Single Cells and Cost Just a Penny Apiece https://singularityhub.com/2025/12/16/these-robots-the-size-of-single-cells-cost-just-a-penny-apiece/
The Future of Focused Research Organizations https://www.essentialtechnology.blog/p/the-future-of-focused-research-organizations
Politics
If you care about happiness, well-being, or growth, you should care about GDP. https://www.vox.com/policy/471950/gross-domestic-product-economics-metrics-growth
A teacher’s teaching quality has very little impact on school achievement (less than 10%). The remaining 90% is due to characteristics associated with students. https://gwern.net/doc/iq/2016-detterman.pdf
Students do not learn more from professors with higher student evaluation of teaching (SET) ratings. New meta-analyses of multisection studies show that SET ratings are unrelated to student learning. https://www.sciencedirect.com/science/article/abs/pii/S0191491X16300323
Teacher effectiveness is negatively correlated with students’ evaluations. https://www.sciencedirect.com/science/article/abs/pii/S0272775714000417
MIT professor and fusion scientist shot dead at his home https://www.nbcnews.com/news/us-news/mit-professor-killed-shooting-home-rcna249582
Ukraine
636.3 “Varshavyanka” (Black Hole) is the pride of the Russian Black Sea Fleet, the “invisible” carrier of the “Calibers” that were used to hit Ukrainian cities. For years, it was presented as a technological fetish and a symbol of control over the Black Sea. After being expelled from Sevastopol, the boat hid in Novorossiysk.
Ukrainian underwater drones did what Russia did not expect at all: they penetrated harbor defenses with an underwater drone in daylight and hacked port surveillance cameras to obtain the footage.
The Black Sea is no longer Russian.
Even in ports.
Note: To date, public satellite imagery shows pier and berth damage. However, it still cannot conclusively demonstrate the extent of damage to the submarine. However, it is plausible that such a detonation so close to the stern could result in a mission kill. The worst damage can be below the waterline: propulsor, rudder, stern planes, pressure hull deformation, mounts, shafts, and seals.