pipette
ESEspañol

Friday, September 25

2,132 new papers appeared. We read every one. These 20 are worth your time.

Each dot is one paper. The pink ones made today's edition. Point at any dot to see its title.

Today's edition

Picked for broad interest, a real step forward and honest claims, with no field allowed to take over.

  1. Somatic differentiation evolves rapidly and repeatedly through the modification of developmental plasticity

    Our results provide the first empirical demonstration that the modification of plastic responses to the environment can drive the evolution of somatic differentiation, a key step in the evolution of a new kind of evolutionary individual.
    AI summary

    In 21 of 48 lines, more colonies showed specialized somatic cells even without cold stress. One lineage developed obligate somatic differentiation, the first published example of this evolution in experiments. Most mutated genes in this lineage had homologs linked to both cold response and somatic cell development. The results show that modifying plastic responses to the environment can drive the evolution of a new kind of individual.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big step
  2. The gene-regulatory evolution of the human skeleton

    These atlases reveal an extensive rewiring of the extracellular matrix (ECM), including a marked suppression of glycosaminoglycan (GAG) biosynthesis, leading to an approximately three-to-fourfold reduction in joint GAG content in humans compared with non-human apes.
    AI summary

    The researchers mapped genetic changes that shaped the human skeleton. They found humans have much less of a joint cushion molecule than other apes. This reduction is likely a key reason humans get more degenerative joint diseases.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Peer-reviewed journalClaims a big step
  3. Agricultural liming is a carbon sink in the Mississippi River Basin

    Here we show, using century-scale records of agricultural liming and anthropogenic acidity inputs for the Mississippi River Basin, that agricultural liming has acted as a net carbon sink over the past century.
    AI summary

    Agricultural liming is the practice of adding carbonate minerals to farmland to reduce soil acidification. The authors analyzed century-long records from the Mississippi River Basin. Better soil pH management may reduce greenhouse gas emissions while improving crop yields and soil health.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Peer-reviewed journalReal-world use
  4. Imaging with GeV muons produced via laser-wakefield-accelerated electrons

    To our knowledge, this represents the first demonstration of imaging dominated by an artificial, laser-driven muon beam.
    AI summary

    The researchers used detectors to measure a muon beam created by shooting laser-accelerated electrons at a solid target. They performed imaging tests with the beam across distances up to 42 meters from the source. Comparisons with computer simulations confirmed the beam contains artificially produced muons with energies of a few GeV. This is the first demonstration of imaging using a laser-driven muon beam, according to the authors.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big step
  5. A near-complete chromosome from a bacterial pathogen integrated into the genome of its arthropod vector

    In addition to numerous NUMTs and diverse families of transposable elements, an almost-complete chromosome of R. africae origin was identified in the cell line genome, localised to the putative sex chromosome.
    AI summary

    Scientists sequenced the genome of a tick cell line and found an almost complete chromosome from the bacterium Rickettsia africae inside it. This large DNA transfer was also found in wild ticks collected from the field. The transferred bacterial DNA could be distinguished from actual infections because plasmid genes were missing. The findings suggest researchers should consider this type of gene transfer when testing vectors for pathogens.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big stepReal-world use
  6. PUBG Ally: A Conversational Embodied Agent as an AI Teammate

    We introduce PUBG Ally, an embodied agent for PUBG: BATTLEGROUNDS that can reason, act autonomously, and play alongside players as a voice-enabled teammate.
    AI summary

    The authors built PUBG Ally, an AI teammate that can talk, think, and play the game PUBG: BATTLEGROUNDS alongside human players. Ally combines a language model that understands speech and makes decisions with a faster system that controls movement, combat, and healing. They used player feedback and preference comparisons to find gaps between test results and what players actually liked. In a live survey across 141 countries, players with confirmed gameplay recommended Ally by a margin of 25.1 percentage points, calling it a teammate or companion.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintReal-world use
  7. Third-generation ultralong-term latent heat storage glass based on a sugar alcohol mixture

    These results establish a third-generation latent heat storage mode, retaining the charged state as a kinetically arrested amorphous solid and thereby fundamentally shifting the stability criterion from supercooled-liquid nucleation suppression to glass-transition and structural-relaxation kinetics.
    Peer-reviewed journalClaims a big stepReal-world use
  8. Spot, Separate, and Enhance: Fully Generative Approach for Audio Mixing

    We introduce Spot, Separate, and Enhance (SSE), the first multimodal, user-guided generative model for audio remixing and enhancement.
    AI summary

    The authors created SSE, a new computer model that remixes and improves audio using both video and text instructions. SSE can rebalance sound, remove unwanted noises, and reduce echo in videos. They built a new dataset called DegradedMix to train and test the model. They used new evaluation methods designed for creative work rather than standard measures. Tests showed SSE works better than existing methods for control and quality.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big stepReal-world use
  9. Residential Electricity Consumption Dataset for Sri Lanka (RECON-SL)

    As the first dataset of its kind in Sri Lanka, and among the few worldwide to link smart meter records with longitudinal survey data at this scale, it provides a unique resource for energy, machine learning and policy research, with known gaps in smart meter coverage documented for users.
    AI summary

    RECON-SL combines electricity records, smart meter data, and household surveys from 4,063 homes in Sri Lanka. Data was collected from October 2022 to January 2025 at monthly, 6-hour, and 15-minute intervals. Smart meter data covers 1,438 households, while surveys gathered information on demographics, housing, and appliances. The dataset contains over 50 million meter readings and more than 11,000 household interviews. It is freely available with documentation and scripts to help researchers reuse it.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintReal-world useEasy to read
  10. Imaging an obstructed Wannier orbital

    By measuring the amplitude and phase of the acoustic pressure field, we map the orbital's -like structure and locate its Wannier center in a void between lattice sites---providing the first phase-resolved image of an obstructed Wannier orbital.
    AI summary

    In normal insulators, electronic charge stays in orbitals centered on atoms. In obstructed atomic insulators, these orbitals sit in empty spaces between atoms instead. The authors built an acoustic metamaterial with a flat band, making Wannier orbitals measurable as energy states. They mapped the orbital's structure and located its center in a void between lattice sites. By changing the lattice edge, they observed the predicted boundary states.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big step
  11. Lu+ optical frequency references with accuracy verified at the 19th digit

    Here we report two 176Lu+ single-ion optical frequency references, each with evaluated fractional frequency uncertainty near 1 × 10−19, which are directly compared by correlation spectroscopy and demonstrate agreement, with a measured relative frequency difference of [−0.1 ± (5.7)stat ± (1.0)sys] × 10−19, where ‘stat’ and ‘sys’ indicate the statistical and systematic uncertainty, respectively.
    AI summary

    The authors built two optical atomic clocks using single lutetium ions. Each clock has a fractional frequency uncertainty near 1 × 10−19. The two clocks agree with a measured relative frequency difference of [−0.1 ± 5.7stat ± 1.0sys] × 10−19. This accuracy will help improve international timekeeping and enable tests of fundamental physics.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Peer-reviewed journalClaims a big step
  12. A prefrontal cortex-hypothalamus circuit for heart rate control in non-human primates

    Here, we identify a prefrontal-hypothalamic circuit that exerts powerful, bidirectional control over heart rate in monkeys.
    AI summary

    The authors identified a brain circuit connecting the prefrontal cortex to the hypothalamus that controls heart rate in monkeys. Stimulating a small prefrontal region called a12 quickly lowered heart rate and increased heart rate variability. This region connects to the lateral hypothalamus, and stimulating that area reproduced the same cardiac effects. Turning off this pathway raised heart rate, showing it normally keeps heart rate in check.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Preprint
  13. A Data-Driven Analysis of Infostealer Malware Victims

    To close this gap, we build a privacy-preserving pipeline that turns illicitly sourced infostealer logs into a reproducible research artifact, minimizing sensitive data while preserving measurement utility, and use it to construct a dataset of 170,298 victims from logs of multiple infostealer families.
    AI summary

    The authors built a privacy-preserving system to study infostealer malware logs from 170,298 victims. They found that gaming and entertainment services were hit especially hard, alongside popular platforms. The data included stolen credentials from law enforcement, government, military, all Ivy League schools, and financial systems. Victims often reused passwords and faced repeated attacks, overlapping with phishing and ransomware targets. They released an anonymized dataset for controlled research access.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Preprint with a published versionClaims a big stepReal-world use
  14. CodeGraph: Open-Taxonomy Knowledge Graph for Source Code with Wikidata Grounding

    We applied our pipeline to the 167 million files of the Stack-Edu corpus, creating the first known large-scale open-taxonomy knowledge graph for source code.
    AI summary

    The authors built a pipeline that uses a code-specialized large language model to annotate source code with semantic labels. These labels are linked to Wikidata, a public knowledge base, through a three-step process including automated queries and an AI research agent. The result is an open knowledge graph called CodeGraph with about 158 million nodes and roughly 1 billion connections. CodeGraph was built from 167 million files in the Stack-Edu corpus and covers 14 programming languages. The authors also created a quality-check method that measures annotation accuracy using human reviewers and AI judges.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Preprint with a published versionClaims a big step
  15. Organic two-dimensional van der Waals heterostructures

    Here we introduce a bottom-up strategy for programmable lattice engineering in organic van der Waals heterostructures of 2D polymers.
    AI summary

    The authors developed a bottom-up method to build organic van der Waals heterostructures using 2D polymers. They stacked layers one by one on water, controlling the lattice alignment, order, and thickness of each layer. Structural tests showed matched lattices in well-aligned systems, while mismatched ones showed moiré patterns and strain distortions. Ultrafast spectroscopy found efficient charge separation at interfaces, and calculations showed built-in electric fields from dipole alignment. Devices showed diode-like rectification ratios above 10^7, which fell as lattice mismatch grew.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Peer-reviewed journalClaims a big step
  16. Lifespan EEG Reference Charts Reveal Aperiodic Confounds in Beta-Band Biomarkers Across Neurological and Neuropsychiatric Disorders

    In this paper, we establish the first comprehensive lifespan reference charts for beta oscillations (N = 22,094, ages: 1-100 years) using complementary analytic approaches that isolate periodic oscillations from aperiodic background activity.
    AI summary

    The authors created the first comprehensive reference charts for beta brain waves using data from 22,094 people aged 1 to 100. They separated true rhythmic beta activity from background aperiodic activity, which had confused previous studies. Beta power followed a three-phase pattern: rising through youth, peaking near age 50, then falling in older age. After adjusting for aperiodic activity, beta frequency showed a previously unknown dip during adolescence around ages 12-13. Most disorders showed reduced beta power, but schizophrenia showed a unique pattern only visible after aperiodic adjustment.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big step
  17. The anthropogenic fingerprint on emerging infectious diseases

    We show that outbreak risks are typically highest in mosaic landscapes where people and livestock live alongside forests and fragmented ecosystems.
    AI summary

    The researchers analyzed 58,319 disease outbreak records across 32 diseases to test how 16 social and environmental factors affect where outbreaks occur. Outbreak risk is highest in mosaic landscapes where people and livestock live near forests and fragmented ecosystems. These factors plus long-term declines in rainfall strongly affect several vector-borne diseases like dengue and Lyme disease. Directly transmitted zoonoses like Ebola and mpox share few common drivers, and most other human pressures vary widely between diseases. Healthcare access most consistently shapes outbreak geography: reporting drops by a median of 32% for each additional hour of travel to a health facility.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Peer-reviewed journalReal-world use
  18. JWST Observations of the Metal-poor T Dwarf WISEA J155349.96+693355.2: The First Brown Dwarf Associated with the Gaia-Enceladus Milky Way Substructure

    The UVW kinematics of WISEA J155349.96+693355.2, its bulk metallicity, and its modest alpha enrichment ([/Fe] = +0.15 dex) are all equivalent to stars identified in the Gaia-Enceladus Milky Way substructure, making it the first brown dwarf with kinematic and chemical evidence of association.
    AI summary

    JWST observed the metal-poor T dwarf WISEA J155349.96+693355.2. The spectrum revealed absorption features indicating a metal-poor low-temperature atmosphere. The dwarf's effective temperature is 960$^{+28}_{-32}$ K and its metallicity is [M/H] = -1.16$^{+0.28}_{-0.25}$ dex. The dwarf's radial velocity and proper motion indicate a highly eccentric Galactic orbit. Its kinematics and metallicity are similar to stars in the Gaia-Enceladus Milky Way substructure.

    Written by Llama 4 Maverick from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big step
  19. AI-Enabled Human Memory Manipulation: Misleading AI-Generated Summaries Distort Human Memory

    People who read a misleading AI summary were significantly less likely to accurately recall the original event, compared to people who read an accurate AI summary.
    AI summary

    The researchers tested whether mistakes in AI-generated summaries change what people remember. They first analyzed AI summaries of videos and found frequent errors, especially missing key details. In an experiment, people watched a car accident video, then read an AI summary with either accurate or wrong information. People who read the misleading summary remembered the original event significantly less accurately. The findings suggest human oversight may not prevent AI errors from distorting memory in high-stakes settings.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintBroad interestReal-world useEasy to read
  20. An Analysis of Streaming Deep Reinforcement Learning for Adaptive Continual Learning in Robotics

    In this paper, we present the first analysis of streaming deep reinforcement learning for adaptive continual learning in robotics.
    AI summary

    Robots often face new situations after training that hurt their performance. The authors studied streaming deep reinforcement learning for continual robot adaptation. After pretraining, this method lets robots adapt to unexpected changes in themselves, their environment, or goals. In quadruped walking tests, stream learning with certain optimizers beat batch methods and improved success by up to 90%. In manipulation tasks, results were only partially successful with some stability and performance limits.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big stepReal-world use