✨ Curiosities

18 articles

🟢 ✨ Curiosities July 18, 2026 · 2 min read

arXiv:2607.15079: BrainPilot Automates Neuroscience Research at a Lower Cost Than SOTA Agent Systems

BrainPilot is a multi-agent system in which a 'principal investigator' agent oversees specialized agents through the entire neuroscience research process, from literature review to analysis and interpretation. It uses a knowledge base of 7,233 sources and 72 methodological units across seven domains, and on Agents' Last Exam and BrainPilotBench-v0 it achieves results comparable to a SOTA framework at lower cost.

🟡 ✨ Curiosities July 16, 2026 · 2 min read

arXiv:2607.13562: AI Advice Suppresses People's Willingness to Say 'I Don't Know' — Even When Wrong

Research shows that AI suggestions nearly eliminate participants' willingness to say 'I don't know,' reducing accuracy to one-third even when AI answers are deliberately wrong.

🟢 ✨ Curiosities July 13, 2026 · 2 min read

Sakana AI: 'Smart Cellular Bricks' Bring Collective Intelligence to the Physical World

Sakana AI has developed modular physical bricks, each running an identical local neural network with no central coordination. Collectively they can recognize the shape of an entire assembly and detect structural damage, extending Sakana's earlier software work on collective intelligence into physical hardware.

🟢 ✨ Curiosities July 7, 2026 · 4 min read

Google Research: Rerouting Less Than 2% of Trips Reduces Urban Congestion and CO2 Emissions

Google Research ran a six-month experiment across 10 US cities: a modified Google Maps algorithm rerouted less than 2% of trips away from selected congested segments. Result — a median +2% speed increase on targeted roads and a 0.5–1% reduction in fuel consumption. The study was published in Nature Cities.

🟢 ✨ Curiosities July 3, 2026 · 4 min read

EO-Agents: Three-Agent LLM Pipeline Generated 160 Scientific Hypotheses from 1,475 NASA Datasets

Researchers developed a three-agent LLM pipeline that uses a NASA Earth Observation Knowledge Graph and a GNN to rank dataset pairs, then automatically generates research hypotheses across glaciology, vegetation phenology, and other domains.

🟢 ✨ Curiosities July 2, 2026 · 4 min read

Two AI Metrics Diverged — Will It Make All the Difference?

Researchers from MIT and Northwestern prove that the conclusion about AI capability democratization or concentration depends entirely on which benchmark you use — not on the actual state of the technology.

🟢 ✨ Curiosities June 23, 2026 · 2 min read

IBM: Wimbledon 2026 gets an AI personal guide for fans and modernized digital platforms

IBM and Wimbledon unveiled new AI-powered fan experiences for The Championships 2026, including personalized interactivity and modernized digital platforms — a continuation of the long-standing partnership that has brought AI to one of the most prestigious sporting events in the world.

🟢 ✨ Curiosities June 20, 2026 · 1 min read

arXiv:2606.20205: Psychological Profiles of LLMs Are Largely a Measurement Artifact, Not a Stable Personality

The paper arXiv:2606.20205 tested 56 instruction-tuned language models with standardized psychological and preference instruments. Using variance decomposition, the authors show that directional response bias explains 81 to 90 percent of differences among models, compared to only 9 to 16 percent in humans, concluding that psychological profiles of models are largely a measurement artifact, not a stable personality.

🟢 ✨ Curiosities June 13, 2026 · 4 min read

arXiv: AI as a hidden third — a study on romantic communication assisted by ChatGPT

Researchers Skyler Wang and Isabella Luppi analyze vernacular discourse from TikTok and Reddit (2023–2026) on the use of AI in romantic communication, introducing the concepts of the covert triad and articulation labor, and concluding that authenticity is being reconfigured from linguistic authorship toward emotional ownership.

🟢 ✨ Curiosities June 4, 2026 · 2 min read

Google Research: passive heart-health monitoring via smartphone camera

Google Research presents a method for passively measuring heart-health indicators using a standard smartphone camera. The goal is accessible and non-invasive health diagnostics without additional hardware. The approach relies on a camera users already have, lowering the barrier to health monitoring.

🟢 ✨ Curiosities June 2, 2026 · 2 min read

arXiv:2606.03883: What does the reasoning structure of large language models actually look like?

Researchers at ETH Zurich have presented a benchmark of logic puzzles and a pipeline that turns reasoning traces into measurable graphs of claims and dependencies. A new metric quantifies reasoning efficiency and reveals differences that accuracy and token count cannot tell apart.

🟢 ✨ Curiosities June 1, 2026 · 2 min read

arXiv:2606.02000: motion control in video diffusion via 3D mesh tokens

A new arXiv study proposes a video diffusion model that conditions human motion control directly on compressed 3D mesh tokens instead of a 2D rendered guide. The method processes video and motion tokens jointly within a transformer, achieving better motion control with fewer artifacts than classic 2D approaches.

🟢 ✨ Curiosities May 29, 2026 · 2 min read

arXiv:2605.30963: AMix-2 Introduces Proteins as a Native Modality in LLMs

AMix-2 is a protein-text foundation model that unifies protein understanding and sequence design in a shared token space. It uses a block-wise diffusion language backbone, introduces the ProteinArena benchmark, and outperforms frontier LLMs while competing with specialized protein models.

🟡 ✨ Curiosities May 23, 2026 · 4 min read

arXiv:2605.22763: AI agent with Lean verification solves 9 open Erdős problems and 44 OEIS conjectures

A team of 20 researchers from DeepMind and MIT CSAIL published the first large-scale evaluation of LLMs for autonomous generation of formal proofs in the Lean theorem prover. The agent combines LLM generation with Lean symbolic verification and autonomously solves 9 of 353 open Erdős problems and proves 44 of 492 OEIS conjectures.

🟢 ✨ Curiosities May 9, 2026 · 2 min read

arXiv:2605.06540: Frontier models fall below diversity threshold in idea generation

When many users employ AI for creative tasks, they all receive similar suggestions — "idea diversity collapse". Researchers introduce an ex ante protocol with an excess-crowding coefficient Δ and diversity ratio ρ. All three tested frontier models fall below the human parity threshold in short stories, marketing slogans, and alternative uses tasks.

🟢 ✨ Curiosities April 27, 2026 · 3 min read

arXiv:2604.21508 BioMiner: multimodal AI extracts protein-ligand bioactivity from literature, 5.59× faster than manual work

The team of Jiaxian Yan and colleagues published on April 23, 2026 BioMiner — a multimodal AI system for automated extraction of protein-ligand bioactivity from scientific literature. The system processes text, tables and molecular structures, achieves F1 0.32 on the new BioVista benchmark (16,457 entries from 500 publications) and in a demonstration application extracts 82,262 data points from 11,683 papers.

🟢 ✨ Curiosities April 23, 2026 · 2 min read

Google Photos Auto Frame uses 3D models and diffusion to expand the frame

Google Photos gained an Auto Frame feature that interprets a 2D photograph as a 3D scene, estimates geometry and camera parameters, then uses latent diffusion models to generate content outside the original frame for alternative compositions.

🟢 ✨ Curiosities April 14, 2026 · 2 min read

ArXiv Camera Artist: Multi-agent AI system that generates video using cinematic language

Researchers have introduced Camera Artist, a multi-agent system that models real filmmaking workflows for narrative video generation. The system coordinates specialized AI agents that simulate the roles of director, cinematographer, and editor for coherent visual storytelling.