Immer mehr ausländische Senioren wandern ins Sozialsystem ein, im Dezember 2025 waren es rund 764.000 Ausländer

Selbstbedienung in der Bananenrepublik D! Leider ist der Artikel nicht eindeutig, ob es sich strickt um nicht EU Ausländer handelt.

Vielleicht wären Quoten für Herkunftsländer oder genauere Prüfung des Asyl bzw. Flüchtlingsstatus und der Einkünfte/Vermögen der ausländischen Bezieher angebracht.

"Im Dezember 2025 bezogen rund 764.000 Personen in Deutschland eine Grundsicherung im Alter – rund 200.000 mehr als noch 2020.

Der Anstieg um 35 Prozent geht nahezu vollständig auf das Konto ausländischer Empfänger, deren Zahl sich in nur fünf Jahren fast verdoppelt hat. Das geht aus aktuellen Zahlen des Statistischen Bundesamtes hervor, die NIUS exklusiv vorliegen."

"... Der Anteil der Ausländer an allen Empfängern von Grundsicherung im Alter stieg damit von 27 Prozent im Jahr 2020 auf 36 Prozent im Jahr 2025. Mit anderen Worten: Inzwischen ist mehr als jeder dritte Bezieher von Grundsicherung im Alter Ausländer. ...

nach Herkunftsländern ... Besonders deutlich fällt hierbei die Entwicklung von zwei Gruppen ins Auge:

Während im Jahr 2020 noch rund 19.500 Ukrainer Grundsicherung im Alter bezogen, waren es im Dezember 2025 bereits 104.285 – eine Verfünffachung. Allein von 2024 auf 2025 kamen rund 9.000 weitere ukrainische Empfänger hinzu.

Auch bei den Top-8-Asylherkunftsländern stieg die Zahl der Empfänger deutlich: von rund 23.900 im Jahr 2020 auf 36.340 im Jahr 2025 – ein Plus von rund 52 Prozent. Die mit Abstand größte Gruppe stellten im Dezember 2025 Syrer mit 14.385 Empfängern, gefolgt von Afghanen (9.170), Iranern (5.105) und Irakern (5.085). Aus der Türkei bezogen im Dezember 2025 weitere 24.000 Personen eine Grundsicherung im Alter. ..."

"Ausländer in Deutschland haben bei Bedürftigkeit im Alter (ab Regelaltersgrenze) oder bei voller Erwerbsminderung Anspruch auf Grundsicherung, sofern sie ihren gewöhnlichen Aufenthalt in Deutschland haben. Sie erhalten Leistungen in gleicher Höhe wie Deutsche, wenn das Einkommen (z.B. eine zu geringe Rente) und das Vermögen nicht zur Sicherung des Lebensunterhalts ausreichen." (Google AI search)

Immer mehr ausländische Senioren wandern ins Sozialsystem ein (newsletter)

Grundsicherung im Alter: Immer mehr ausländische Senioren wandern ins Sozialsystem ein

London Marathon 2026 results: Sabastian Sawe makes history with first competitive sub-two-hour marathon

Wow! Very impressive!

"... The 31-year-old Kenyan crossed the line to win in one hour 59 minutes 30 seconds, more than one minute faster than the late [Kenyan] Kelvin Kiptum's previous record of 2:00:35, set in 2023. ..."

London Marathon 2026 results: Sabastian Sawe makes history with first competitive sub-two-hour marathon - BBC Sport

Source

On Image Generators are Generalist Vision Learners

This could be an interesting, new paper by Kaiming He & Thomas Funkhouser and their team!

From the abstract:

"Recent works show that image and video generators exhibit zero-shot visual understanding behaviors, in a way reminiscent of how LLMs develop emergent capabilities of language understanding and reasoning from generative pretraining. While it has long been conjectured that the ability to create visual content implies an ability to understand it, there has been limited evidence that generative vision models have developed strong understanding capabilities.

In this work, we demonstrate that image generation training serves a role similar to LLM pretraining, and lets models learn powerful and general visual representations that enable SOTA performance on various vision tasks.

We introduce Vision Banana, a generalist model built by instruction-tuning Nano Banana Pro (NBP) on a mixture of its original training data alongside a small amount of vision task data.

By parameterizing the output space of vision tasks as RGB images, we seamlessly reframe perception as image generation. Our generalist model, Vision Banana, achieves SOTA results on a variety of vision tasks involving both 2D and 3D understanding, beating or rivaling zero-shot domain-specialists, including Segment Anything Model 3 on segmentation tasks, and the Depth Anything series on metric depth estimation.

We show that these results can be achieved with lightweight instruction-tuning without sacrificing the base model's image generation capabilities.

The superior results suggest that image generation pretraining is a generalist vision learner. It also shows that image generation serves as a unified and universal interface for vision tasks, similar to text generation's role in language understanding and reasoning.

We could be witnessing a major paradigm shift for computer vision, where generative vision pretraining takes a central role in building Foundational Vision Models for both generation and understanding."

[2604.20329] Image Generators are Generalist Vision Learners

Renewables Overtook Global Electricity Demand Last Year. Really!

I bet, we are dealing here with a lot of wishful thinking and data manipulation! More propaganda than fact?

Was the solar+wind power generated electricity actually used?

The vague language of the news reports does not distinguish capacity from generation or private household generation from commercial generation etc. I have some serious doubts about the claims of such enormous solar power generation growth as stated by Ember!

What exactly is included in the euphemistically and falsely called "clean energy"?

Who funds the think tank Ember?

Ember vision/mission statement: "We’re a global energy think tank that aims to accelerate the clean energy [???] transition with data and policy"

"“Record growth in solar, especially in China and India, was a driving factor for clean energy sources [???] surpassing the world’s strong demand for electricity in 2025, according to a new global power analysis.

Clean power generation [???] grew 887 terawatt hours last year, exceeding overall global electricity demand growth of 849 terawatt hours, according to a report by energy think tank Ember, released after midnight Tuesday London time.”"

"... Together with growth in other clean sources, this solar surge drove clean power to meet all global electricity demand growth in 2025. Solar alone met three-quarters (75%) of the increase, while solar and wind together met almost all of it (99%).

In total, clean generation rose by 887 TWh, slightly exceeding demand growth of 849 TWh. As a result, fossil generation fell by 0.2%, making 2025 only the fifth year this century without growth in fossil electricity. ..."

Renewables Overtook Global Electricity Demand Last Year - Human Progress

Renewable energies overtook global electricity demand last year, led by solar growth in China, India

Solar surge halts fossil electricity growth worldwide in 2025 (original news release)

One of cholera’s great enemies is found in the human gut, a bacteriophage

Recommendable!

"... found that in the Ganges Delta, cholera bacteria rapidly gain and lose special armour that protects against attacks from the virus, known as bacteriophage ICP1.

The new research ... highlighted that maintaining these anti-viral defences leads to lower disease severity of cholera in humans and reduced ability to spread outside the country for this bacterial strain. ...

By looking at the ecology of cholera in South Asia, this study challenges the long-held belief that the Ganges Delta is the global source of cholera. Knowing more about the strains and the factors that influence the spread of cholera bacteria in different regions could help provide an early warning system, identifying high-risk strains before they escalate and allowing for early intervention. ...

Globally, we are in the seventh cholera pandemic, which started in 1961, with an estimated 1.3 to 4 million cases and up to 143,000 deaths per year from the condition worldwide. It has been shown that the seventh pandemic is caused by V. cholerae strain 7PET O1, originating from the Bay of Bengal, which borders Bangladesh and India, and it was thought that the Ganges Delta was the global source of cholera.

This new research sequenced bacterial samples from across Bangladesh and North India, creating the most comprehensive dataset of cholera in this area to date, containing over 2,300 genomes collected across approximately 20 years. They found that it was the Ganges Basin, not the Ganges Delta, that was the primary global source of cholera in that time.

By tracking the bacterial spread, they also uncovered that the bacteria do not simply follow the flow of rivers. Instead, they tend to stay within national borders, suggesting that human travel and population density are more important for cholera transmission than the natural environment.

They also found V. cholerae in Bangladesh, strain 7PET O1, rapidly gain and lose genetic elements known as defence systems, which act like armour helping them survive against their viral nemesis, the bacteriophage ICP1. ..."

From the abstract:

"The seventh pandemic of cholera, caused by the seventh pandemic El Tor lineage of Vibrio cholerae, was previously shown to have emanated in three global waves from the Bay of Bengal, bordering Bangladesh and India.

However, the respective roles of the Ganges Delta and Basin regions in seeding these global pandemic waves were not known.

Here we show that, although transmission events occur between Bangladesh and India, V. cholerae in the two countries has largely evolved separately over the past 20 years, apparently constrained by national borders rather than by hydrological features, such as the Ganges Delta and Basin.

Evolution within Bangladesh was distinct from that seen in India, involving rapid gain and loss of genes and mobile genetic elements, particularly those involved in phage defence. The loss of these systems was associated with increased risk of severe disease and transmission outside Bangladesh.

Lineage replacement in Bangladesh in 2018, resulting in a major change in phage defence systems, was accompanied by a rapid change in the lineage and anti-defence system of lytic phage ICP1.

Here we show that the Ganges Basin, falling across Bangladesh and Northern India, rather than the Ganges Delta, probably acts as a global launch pad for pandemic disease. This shifts our understanding of Bangladesh as the purported global source of cholera and underscores the potential role of phage in controlling spread of lineages within the current seventh pandemic."

One of cholera’s great enemies is found in the human gut "Cholera-causing bacteria are locked in an evolutionary arms race with a viral nemesis, according to a new genomic study."

Evolution of pandemic cholera at its global source (open access)

Fig. 1: Dynamics of V. cholerae sublineages in Bangladesh and their genetic profiles over time.

Pure classical physics can explain quantum phenomena, study shows

Amazing stuff!

"A reformulation of the classical Hamilton-Jacobi equation, incorporating density and multiple least-action paths, can exactly reproduce quantum phenomena such as the double-slit experiment, quantum tunneling, and hydrogen atom wave functions. This approach mathematically bridges classical and quantum mechanics, showing that quantum behavior can be computed using classical principles without approximations."

" ... MIT scientists have now shown that certain mathematical ideas from everyday classical physics can be used to describe the often weird and nonintuitive behavior that occurs at the quantum, subatomic scale.

In a paper appearing today in the journal Proceedings of the Royal Society A Mathematical Physical and Engineering Science, the team shows that the motion of a quantum object can be calculated by applying an idea from classical physics known as "least action." With their new formulation, they show they can arrive at exactly the same solution as the Schrödinger equation—the main description of quantum mechanics—for a number of textbook quantum-mechanical scenarios, including the double-slit experiment and quantum tunneling. ..."

From the abstract:

"We show that the Schrödinger equation can be solved exactly based only on classical least action.

Fundamental postulates of quantum mechanics can in turn be derived directly from this construction. The results extend to the relativistic Klein-Gordon, Pauli, and Dirac equations, and suggest a smooth transition between physics across scales. 

Most quantum mechanics problems have classical versions which involve multiple least action solutions. The associated classical multipaths stem either from the initial position or momentum distribution, or from branch points, generated, e.g. by a multiply connected manifold (double slit experiment), by spatial inequality constraints (particle in a box), or by a singularity (Coulomb potential). We show that the exact Schrödinger wave function  can be constructed by combining this classical multi-valued action with the classical density ⁠, computed analytically from  along each extremal action path.

The construction is general and does not involve any semi-classical approximation.

Quantum wave collapse at measurement can be derived from the classical density change. Entanglement corresponds to a sum of classical particle actions mapping to a tensor product of spinors. The results also provide a simpler computational alternative to Feynman path integrals, as they use only a minimal subset of classical paths."

Classical physics can explain quantum weirdness, study shows

On computing quantum waves exactly from classical action (open access)

New study bridges the worlds of classical and quantum physics (original news release) "The weird quantum behavior of subatomic particles can be understood through everyday classical ideas, MIT researchers show."

On computing quantum waves exactly from classical and relativistic action (prepint, first published 5/10/2024, open access)

China: A street cleaning truck in action

As seen in Zhengzhou, Henan province.

It looks like a hair dryer and it sprays water into the air like a shower head!

China: A casino for elderly citizens inside an underground subway station

On my latest visit to Zhengzhou, Henan province, I noticed two tables in a side wing of an underground subway station. There were two tables at least 8 elderly individuals sitting around these two tables, playing a card game together at each table.

I thought this was curious!

P.S. I don't think they were card playing for money.

China: It rains and you forgot your umbrella. No problem!

As seen in Zhengzhou, Henan province.

This umbrella rental machine is just for you! 😊 So far I have seen several of these rental machines.

China: Highway to flower garden

On my latest visit to Zhengzhou, Henan province, China I noticed that a lot of the many highway overpasses in this city were decorated with flowers along the left and right edge of the highway. 

It looked very nice like a flower garden!

 

China: Some progress with ticket vending machines for public transportation in an urban area

As seen in Zhengzhou, Henan province.

Very nice, most of the functions of this vending machine were translated into english.

However, if you e.g. have to look up the different routes of the subway system it is all still in chinese language.

Disclaimer

Since end of February, I  am blogging from behind the Great Firewall of China.

My Internet service in China is very spotty. Thus, I am not able to blog as usual.

Overlooked Brain Connections Hold Clues to Cognition and Mental Health

Amazing stuff!

"Key points

• Scientists who use imaging to understand the brain’s complexity often focus on the strongest signals and discard the rest.
• A new study reveals that connections routinely overlooked as “noise” during neuroimaging data analysis can predict behavior with remarkable accuracy.
• The finding could help explain why some people with psychiatric illness don’t respond to treatments, and it could identify new targets for therapeutics.

...

For the study, researchers investigated whether signals discarded by feature selection could reveal meaningful insights about brain and behavior. The team examined brain imaging and behavioral data from more than 12,000 participants across four major U.S. datasets. For every participant, the team calculated the strength of association between brain connections and the outcome they wanted to predict.

All the connections were then ranked from the strongest to weakest associated and divided into 10 non-overlapping groups.

Group one contained the top 10% of connections, those that scientists usually select, while groups two through 10 held the remaining 90% of connections—the connections often dismissed as noise. The team then built 10 prediction models, one for each group. ...

The team found that lower-ranked connections—groups two through nine—consistently achieved prediction accuracy similar to the top 10% of connections. 

In some cases, models built on lower groups of connections performed better than those trained on the top group. The authors suggest this might be because predictive information is widely distributed throughout brain connections and not just concentrated within the strongest ones. ..."

From the abstract:

"A central objective in human neuroimaging is to understand the neurobiology underlying cognition and mental health.

Machine learning models trained on neuroimaging data are increasingly used as tools for predicting behavioural phenotypes, enhancing precision medicine and improving generalizability compared with traditional MRI studies.

However, the high dimensionality of brain connectivity data makes model interpretation challenging. Prevailing practices rely on selecting features and, implicitly, interpreting identified feature networks as uniquely representative of a given phenotype while overlooking others. Despite its widespread use, how univariate feature selection balances the trade-off between simplification for optimizing modelling and oversimplification that misrepresents true neurobiology remains understudied.

Here, using four large-scale neuroimaging datasets spanning over 12,000 participants and 13 outcomes, we demonstrate that edges discarded by feature selection can achieve significant prediction accuracies while yielding different neurobiological interpretations. These results are observed across cognitive, developmental and psychiatric phenotypes, extend to both functional connectivity (functional MRI) and structural connectomes (diffusion tensor imaging) , and remain evident in external validation.

They suggest that focusing on only the top features may simplify the neurobiological bases of brain–behaviour associations. Such interpretations present only the tip of the iceberg when certain disregarded features may be just as meaningful, potentially contributing to ongoing issues surrounding reproducibility within the field. More broadly, our results reinforce that subtle brain-wide signals should not be ignored."

Overlooked Brain Connections Hold Clues to Cognition and Mental Health | Yale School of Medicine

Feature selection leads to divergent neurobiological interpretations of brain-based machine learning biomarkers (open access)

Fig. 1: CPM [connectome-based predictive modelling ] across non-overlapping decile-ranked brain connectivity features.

a, Workflow illustrating the decile-based CPM pipeline, including the initial correlation of connectivity features with phenotypic outcome, ranking features on the basis of group-level correlations between edges and phenotype, splitting features into deciles, and evaluating each decile-based model. DTI, diffusion tensor imaging.

b, Violin plot showing the predictive performance of models trained on each decile of features within the PNC dataset for executive function.

c–e, Radar plots depicting predictive performance across deciles for PNC executive function (c, left), PNC language abilities (c, right), HCPD executive function (d, left), HCPD language abilities (d, right), HBN executive function (e, left) and HBN language abilities (e, right). Bold decile numbers indicate significant predictions.

This 2,200-year-old Roman wreck hid a repair story that rewrites how ancient ships survived long voyages

Amazing stuff! The abstract of this new research paper is disappointing! It is not even an abstract, but more of a teaser.

This paper even uses what appears to be computer vision to generate images presented in this paper! See below.

"Ever since humans have embarked on sea voyages, they needed to ensure vessels were waterproof, resistant to salty seawater, and could withstand microorganisms or sea-dwellers like worms. Until the mid-20th century, however, the study of non-wood materials used to build ships was overlooked. Even today, little work has been done on materials used for waterproofing.

Now, in a new ... study, researchers ... have examined the protective coating of the Roman Republic shipwreck Ilovik–Paržine that sank around 2,200 years ago off the coast of what is now Croatia. ...

"Studying the coatings, we found two different kinds on this vessel: one made of pine tar, also called pitch, and the other of a mixture of pine tar and beeswax. Analysis of pollen in the coating made it possible to identify the plant taxa present in the immediate environment during the construction or repairs of the ship." ...

The wreck was discovered in 2016 and since then the ship itself and its cargo has been examined multiple times. The current study, however, is the first to combine pollen and molecular analyses to characterize the ship's coating and vegetation present during its production and application on the hull. ..."

From the abstract:

"Introduction: The construction of a vessel (from a boat to a large ship) and its maintenance requires waterproofing of its hull and protection against water corrosion and the aggression of microorganisms, worms and other pests. What could be more logical than using an easily accessible and applicable hydrophobic adhesive material?

Many substances have been used over time such as resins, bitumen, plant tars, pure or mixed with beeswax, fats, inorganic elements. Pliny the Elder already mentions zopissa, a mixture of pitch and beeswax (Natural History XVI, 23).

The strong expansion of shipbuilding between the 13th and 19th centuries generated a veritable industry of plant tars.

Methods: In this research work, a new interdisciplinary approach involving the combined use of molecular, palynological and statistical indicators has been implemented to characterize ancient waterproofing materials. This analytical strategy opens new fields of investigation in naval archaeology.

Results and discussion: Beyond the characterization of materials (nature, manufacturing processes, naval techniques, degree of alteration), it especially reveals information about the surrounding vegetation during the production or the application of the waterproofing material. This approach has been applied to the study of the protective coating of the Roman Republican wreck Ilovik–Paržine 1 (around the middle of the second c. BC) found in Paržine Bay (Ilovik Island, Croatia)."

This 2,200-year-old Roman wreck hid a repair story that rewrites how ancient ships survived long voyages

Adhesive coatings in naval archaeology: molecular and palynological investigations on materials from the Roman Republican wreck Ilovik–Paržine 1 (Croatia) (open access)

Fig. 1 View of the excavation of the bow area of the Ilovik-Paržine 1 shipwreck. In the foreground, the cargo of logs and amphoras can be seen. Archaeologists are working near the structure of the bow complex 

A New Type of Neuroplasticity Rewires the Brain After a Single Experience

Recommendable!

"... Recently, neuroscientists described a new form of neuroplasticity that might be helping the brain learn across a timescale of several seconds — long enough to capture the behavioral process of learning from a single experience. In two recent reviews ... describe “behavioral timescale synaptic plasticity,” or BTSP. This type of learning in the hippocampus, the brain’s memory hub, is caused by an electrical change that affects multiple neurons at once and unfolds across several seconds. Researchers suspect that it may help the brain learn in a single attempt. ..."

From the abstract (1):

"Understanding how brains learn and remember remains among the most important challenges in science. Recent studies in the hippocampus implicate a new form of synaptic plasticity, named behavioral timescale synaptic plasticity (BTSP), in the generation of experience-based learning and memory. BTSP is a strong, bidirectional type of plasticity that affects synaptic weights over many seconds of time.

It is induced by single dendritic plateau potentials, as opposed to many action potentials, and is thus capable of producing new place cells in one trial.

Plateau potential initiation is controlled, at least in part, by local feedback inhibition and an instructive input from a higher-order brain region that potentially links the plasticity to current experience.

The new credit assignment procedure in BTSP provides a nonstandard mechanism for memory storage and retrieval that could mitigate the need for widespread synapse stabilization. In addition, it may allow hippocampal networks both to form memories of specific behavioral episodes and to generalize on the basis of past episodes.

Finally, recent BTSP investigations could provide a basis for future explorations into how brains learn and remember, ranging from the systems and cognitive levels down to the basic biochemical building blocks of learning and memory."

From the abstract (2):

"Hebbian synaptic plasticity is currently the main framework to relate neuronal activity, network structure, and learning and memory.

However, recent experimental and computational modeling studies have revealed a new form of synaptic plasticity termed behavioral timescale synaptic plasticity (BTSP).

It is triggered by dendritic plateau potentials associated with somatic burst firing, causes large changes in synaptic strength in a single shot, and operates on the timescale of seconds.

Here we review the recent advances in our understanding of the circuit, cellular, and molecular mechanisms of BTSP, its prevalence in the brain, its role in shaping neuronal representations, and the emerging ideas regarding its contribution to different forms of learning."

A New Type of Neuroplasticity Rewires the Brain After a Single Experience | Quanta Magazine "“Neurons that fire together, wire together” is not the full story. A novel mechanism explains how the brain can learn across longer timescales."

Behavioral timescale synaptic plasticity: properties, elements and functions (1, review article, no public access)

Behavioral Timescale Synaptic Plasticity: A Burst in the Field of Learning and Memory (2, no public access)

Dendrites, the extended branches that receive signals from other neurons, are the star players in a recently described type of neuroplasticity. In this image of stained pyramidal neurons from the cerebral cortex, rootlike dendrites extend from the cell bodies.

HAQERs evolved after hominins split from chimps but before Homo sapiens diverged from Neanderthals

Amazing stuff!

"... Researchers built on past studies documenting the language abilities—and saliva—of 350 elementary school children. They wanted to analyze genetic regulatory sequences called Human Ancestor Quickly Evolved Regions (HAQERs), which influence how genes get expressed. The students’ DNA helped the team confirm HAQERs’ importance in processing and demonstrating language. Then, the researchers looked for the presence of HAQERs in regions of DNA known to have evolved in ancient primates and hominins.

They found that HAQERs evolved after hominins split from chimps but before Homo sapiens diverged from Neanderthals—meaning complex communication likely preceded our own species. This “sliver of the genome has remained relatively constant, even as other aspects have been going up and up and up to make modern humans smarter and smarter ,” author Jacob Michaelson said in a statement. “We can say humans at least had the ‘hardware’ for language earlier than what we previously thought.”

As for why the HAQERs didn’t continue to evolve much after the Neanderthal-human split, the researchers suggest it’s because they promote fetal development—and give babies bigger heads. That tradeoff would have quickly cost the lives of ancient mothers and infants."

From the abstract:

"Language is a defining feature of our species, yet the genomic changes enabling it remain poorly understood. Despite decades of work since FOXP2’s discovery, we still lack a clear picture of which regions shaped language evolution and how variation contributes to present-day phenotypic differences.

Using an evolutionary stratified polygenic score approach, we find that human ancestor quickly evolved regions (HAQERs) are associated with spoken language abilities (discovery N = 350, total replication N > 100,000).

HAQERs evolved before the human-Neanderthal split, giving hominins increased binding of Forkhead and Homeobox transcription factors, and show evidence of balancing selection across the past 20,000 years.

Language-associated variants in HAQERs appear more prevalent in Neanderthals, and HAQER-like sequences show convergent evolution across vocal-learning mammals. Our results reveal how ancient innovations continue shaping human language."

ScienceAdviser

Ancient regulatory evolution shapes individual language abilities in present-day humans (open access)

Fig. 1. Overview of this study and key findings. (ka, thousand years ago; Ma, million years ago.)

Sperm cargo is richer and more important than we thought

Amazing stuff!

I remember way back then when some feminists and some female women's rights activists mocked/dismissed men as mere, dispensable sperm providers. 

"Sperm have long been thought of as streamlined DNA delivery vehicles, carrying little more than a father’s genes to the egg. But a new study shows that in mice, sperm may transmit the father’s influence in another way: During their passage through the epididymis, the coiled tube where they mature after leaving the testes, sperm pick up messenger RNAs (mRNAs)—RNA transcripts of genes that contain the genetic instructions for making proteins. And these mRNAs seem to be transferred to the fertilized egg, a team reports this week in Nucleic Acids Research.

Researchers already knew sperm ferry small RNAs, RNA fragments that can silence gene expression and have been implicated in transmitting the effects of paternal diet, stress, and  exercise to offspring. But mRNAs could be a far more direct route for paternal influence ... The work doesn’t prove these mRNAs from sperm actually function in embryos, she noted, but the discovery of unexpected cargo is “potentially very significant.”"

"... The researchers also found that some mRNAs present in mature sperm are absent from unfertilized eggs, but appear in zygotes after fertilization—suggesting sperm deliver these transcripts to the embryo, ... Because environmental conditions can trigger mRNA production in the father, the observed mRNA transfer “really establishes a mechanism for how the environment can directly influence sperm to then potentially influence the next generation,” ...

the researcher injected long RNA sequences into parthenotes—mouse eggs triggered to divide and develop without sperm. They found the injections shifted the cells’ gene expression to resemble normally fertilized embryos. (They team used RNAs longer than 200 nucleotides, but not mRNA ... The results suggest large RNAs from sperm “can do something after fertilization to regulate embryonic gene expression,” ...

The mechanism may not be limited to mice. The team also sequenced mRNAs from mature human sperm and found counterparts to many mRNAs in mouse sperm, suggesting humans may also deliver RNA messages to their offspring this way."

From the abstract:

"The epididymis plays a critical role in promoting sperm maturation, including remodeling the sperm RNA payload. While small RNAs have been extensively studied in this context, the epididymal contribution to larger sperm RNAs, such as messenger RNAs (mRNAs), remains underexplored. This is largely due to the translational quiescence of mature spermatozoa and the hypothesis that these RNAs are residual by-products of spermatogenesis.

Yet, mRNAs carried by sperm have been detected in the zygote, indicating they may act beyond fertilization. However, whether epididymal somatic cells contribute mRNAs to sperm, as they do small RNAs, has not been experimentally examined. Here, we provide a comprehensive analysis of the mRNA landscape of mouse sperm, epithelial cells, and extracellular vesicles (EVs) isolated from the proximal (caput) and distal (cauda) epididymis.

Through this analysis and sperm-EV co-incubation experiments, we demonstrate the transfer of mRNAs from epididymal EVs to sperm.

Further, through sperm RNA microinjection into zygotes, we uncover gene regulation in the early embryo driven by the introduction of sperm RNAs, specific to >200-nucleotide RNA species.

These findings reveal the dynamic mRNA profile of sperm that is delivered to the egg and demonstrate that RNA species beyond small RNAs are capable of influencing preimplantation embryo gene expression."

ScienceAdviser

Sperm carry unexpected genetic messages "Maturing mouse sperm get loaded with full-length messenger RNAs that are transferred to fertilized egg, suggesting a new route for paternal influence"

Epididymal extracellular vesicles harbor and convey mRNA to sperm for transfer to zygotes (open access) [This is a long research article!]

Graphical abstract

On Language models recognize dropout and Gaussian noise applied to their activations

This could be an interesting new paper by Yoshua Bengio and his team!

Caveat: I have not yet read the paper.

From the abstract:

"We provide evidence that language models can detect, localize and, to a certain degree, verbalize the difference between perturbations applied to their activations. More precisely, we either

(a) mask activations, simulating dropout, or

(b) add Gaussian noise to them, at a target sentence.

We then ask a multiple-choice question such as ``Which of the previous sentences was perturbed?'' or ``Which of the two perturbations was applied?''.

We test models from the Llama, Olmo, and Qwen families, with sizes between 8B and 32B, all of which can easily detect and localize the perturbations, often with perfect accuracy. These models can also learn, when taught in context, to distinguish between dropout and Gaussian noise. Notably, qwenb's zero-shot accuracy in identifying which perturbation was applied improves as a function of the perturbation strength and, moreover, decreases if the in-context labels are flipped, suggesting a prior for the correct ones -- even modulo controls.

Because dropout has been used as a training-regularization technique, while Gaussian noise is sometimes added during inference, we discuss the possibility of a data-agnostic ``training awareness'' signal and the implications for AI safety."

 

[2604.17465] Language models recognize dropout and Gaussian noise applied to their activations

On Back into Plato's Cave: Examining Cross-modal Representational Convergence at Scale

Food for thought! This could be an interesting paper by Alexei A. Efros and his team.

This paper raises some critical issues!

Caveat: I have not yet read the paper.

From the abstract:

"The Platonic Representation Hypothesis suggests that neural networks trained on different modalities (e.g., text and images) align and eventually converge toward the same representation of reality.

If true, this has significant implications for whether modality choice matters at all. We show that the experimental evidence for this hypothesis is fragile and depends critically on the evaluation regime.

Alignment is measured using mutual nearest neighbors on small datasets (1K samples) and degrades substantially as the dataset is scaled to millions of samples.

The alignment that remains between model representations reflects coarse semantic overlap rather than consistent fine-grained structure.

Moreover, the evaluations in Huh et al. are done in a one-to-one image-caption setting, a constraint that breaks down in realistic many-to-many settings and further reduces alignment.

We also find that the reported trend of stronger language models increasingly aligning with vision does not appear to hold for newer models. Overall, our findings suggest that the current evidence for cross-modal representational convergence is considerably weaker than subsequent works have taken it to be. Models trained on different modalities may learn equally rich representations of the world, just not the same one."

[2604.18572] Back into Plato's Cave: Examining Cross-modal Representational Convergence at Scale

On Separating Geometry from Probability in the Analysis of Generalization

This could be an interesting new paper by Benjamin Recht and his coauthor!

Caveat: I have not yet read this paper.

From the abstract:

"The goal of machine learning is to find models that minimize prediction error on data that has not yet been seen. Its operational paradigm assumes access to a dataset S and articulates a scheme for evaluating how well a given model performs on an arbitrary sample. The sample can be S (in which case we speak of ``in-sample'' performance) or some entirely new S' (in which case we speak of ``out-of-sample'' performance). Traditional analysis of generalization assumes that both in- and out-of-sample data are i.i.d. draws from an infinite population.

However, these probabilistic assumptions cannot be verified even in principle.

This paper presents an alternative view of generalization through the lens of sensitivity analysis of solutions of optimization problems to perturbations in the problem data. Under this framework, generalization bounds are obtained by purely deterministic means and take the form of variational principles that relate in-sample and out-of-sample evaluations through an error term that quantifies how close out-of-sample data are to in-sample data.

Statistical assumptions can then be used ex post to characterize the situations when this error term is small (either on average or with high probability)."

[2604.19560] Separating Geometry from Probability in the Analysis of Generalization

Debatte um Absetzung: Selbst alte Weggefährten halten Trump für wahnsinnig. Wirklich!

Auch unter deutschen Journalisten grassiert das Trump Derangement Syndrome! Wie z.B. Sofia Dreisbach, Washington!

Debatte um Absetzung: Selbst alte Weggefährten halten Trump für wahnsinnig | FAZ (behind paywall) "Sogar im MAGA-Lager wird die Frage lauter: Ist der Präsident noch zurechnungsfähig? Trump aber denkt nicht daran, einen Gang zurückzuschalten."

Meet GitNexus: An Open-Source MCP-Native Knowledge Graph Engine That Gives Claude Code and Cursor Full Codebase Structural Awareness

Recommendable!

"GitNexus is an open-source knowledge graph engine that indexes any codebase into a structured dependency map — capturing every function call, import, class inheritance, and execution flow using Tree-sitter AST parsing — and exposes it to AI coding agents like Claude Code, Cursor, Codex, OpenCode, and Windsurf via a Model Context Protocol (MCP) server. Instead of letting agents edit code blind and ship breaking changes, GitNexus pre-computes the entire dependency structure at index time so agents can answer architectural questions like "what depends on this function?" in a single query, with confidence-scored blast radius analysis, 360-degree symbol context, pre-commit impact detection, and coordinated multi-file renames — all triggered by one command: npx gitnexus analyze. Fully local, zero server, 13 languages supported, and already at 19,100 GitHub stars"

Meet GitNexus: An Open-Source MCP-Native Knowledge Graph Engine That Gives Claude Code and Cursor Full Codebase Structural Awareness - MarkTechPost

DeepSeek AI Releases DeepSeek-V4

Recommendable! Seems the latest DeepSeek model comes with several  interesting innovations!

"Key Takeaways

• Hybrid CSA and HCA attention cuts KV cache to 10% of DeepSeek-V3.2 at 1M tokens.
• Manifold-Constrained Hyper-Connections (mHC) replace residual connections for more stable deep layer training.
• The Muon optimizer replaces AdamW for most parameters, delivering faster convergence and training stability.
• Post-training uses On-Policy Distillation from 10+ domain experts instead of traditional mixed RL.
• DeepSeek-V4-Flash-Base outperforms DeepSeek-V3.2-Base despite having 3x fewer activated parameters.

..."

DeepSeek AI Releases DeepSeek-V4: Compressed Sparse Attention and Heavily Compressed Attention Enable One-Million-Token Contexts

US Defense Secretary Hegseth warns European allies to stop 'free riding' and help reopen the strait

Good news! Bravo!

During the Cold War, the European NATO members were accused like this many times!

After the Cold War it is perhaps time to dissolve the NATO. Let the Europeans build their own defense!

Hegseth warns European allies to stop 'free riding' and help reopen the strait (partially behind paywall)

China: Teeth ceaning at a dental clinic in Zhengzhou

What an experience!

You pay per natural tooth, crowns are not counted! I never encountered this before in my life.

Dental flossing not applied!

No mirror was provided at the end of service to review your cleaned teeth!

The ultrasonic scaler was applied extensively before and after polishing. Why?

There were two patient seats, two patients in the same small room with about six dental staff. 

Hidden mutations in immune cells linked to autoimmune disease

Good news! This could be a breakthrough!

"New research suggests that autoimmune diseases may be driven by DNA mutations in immune cells that remove the natural brakes on the immune system. It reveals a previously hidden role for somatic mutations — DNA changes acquired throughout life — in diseases beyond cancer.

Researchers ... used a series of cutting-edge techniques to identify previously unseen changes in DNA that may contribute to thyroid autoimmunity, where the immune system attacks the thyroid gland. ...

The researchers used several advanced DNA analysis techniques.
Firstly, they used a method called NanoSeq, which they recently developed and allows detection of rare mutations invisible to traditional DNA sequencing methods, to look for genetic changes that may drive the disease. They found that many B cells had developed inactivating mutations in key genes that normally control the immune system.

Next, using additional methods that look at the DNA of individual cells and microscopic areas of tissue, the researchers found that many B cells in each patient carried several mutations in key genes.

Two critical immune-checkpoint genes, TNFRSF14 and CD274 (or PDL1), were often lost independently in multiple clones of mutated B cells in each patient. Some of these clones had even acquired as many as six driver mutations over many years, silently building up changes in DNA before symptoms appeared, a highly unexpected observation outside of cancer.

Importantly, artificial inactivation of these genes, in experimental studies or during cancer immunotherapy, is known to cause thyroid autoimmunity. The researchers have now found frequent mutations in these genes occurring in autoimmune patients. ..."

From the abstract:

"Our immune system contains multiple checkpoints to prevent the activation of self-reactive lymphocytes. How some lymphocytes escape these constraints to cause autoimmune disease remains poorly understood. A long-standing hypothesis posits that somatic mutations in immune-regulatory genes may enable self-reactive lymphocytes to bypass tolerance checkpoints1–3, but testing this has been challenging due to technical limitations.

Here, we use whole-exome and targeted NanoSeq, an accurate single-molecule DNA sequencing protocol, to comprehensively search for driver mutations in autoimmune thyroid disease. This revealed many B cell clones convergently acquiring loss-of-function mutations in the key immune checkpoint genes TNFRSF14 (HVEM) and CD274 (PD-L1), as well as less frequent mutations in other immune genes.

In highly inflamed biopsies, we detected tens to hundreds of independent immune checkpoint mutant clones. Laser microdissection, methylation sequencing, spatial transcriptomics, immunostaining, single-nucleus DNA sequencing, and antibody synthesis localised these mutations to B cells, confirmed some to be self-reactive, and identified clones carrying multiple hits.

We found widespread TNFRSF14 biallelic loss, and clones with as many as 4-6 driver mutations. Whilst each clone accounts for a small fraction of cells (typically <1%), the myriad mutant clones in each donor amounted to a substantial fraction of B cells harbouring driver mutations.

Our results support the hypothesis that somatic mutations in autoimmune lymphocytes may allow them to escape tolerance constraints through a polyclonal cascade of somatic evolution, providing new insights into the molecular basis of autoimmune disease."

Hidden mutations in immune cells linked to autoimmune disease "Mutations in immune cells may be the missing piece in the autoimmune disease puzzle."

Polyclonal selection of immune checkpoint mutations in thyroid autoimmunity (no public access)

Academic freedom slides globally, as it takes a nosedive in the US. Really!

Let me guess, a case of Trump Derangement Syndrome by the Royal Society of Chemistry! What a piece of junk journalism!

"Academic freedom continues to decline globally, with a marked fall in the US, according to the 2026 update to the Academic Freedom Index. The index revealed that over the last decade academic freedom has slipped in 50 countries, with just nine registering improvement. ...

‘This academic freedom erosion is part of broader democratic backslide that we’re seeing in the US,’ [???] says Jennifer Jones, director of the Center for Science and Democracy at the Union of Concerned Scientists. ‘It mirrors the rise of authoritarianism that we’ve seen in Hungary and Turkey [???], where higher education is often targeted.’ ..."

Academic freedom slides globally, as it takes a nosedive in the US | News | Chemistry World

The Trump administration said it would allow executions by firing squad and restore the use of lethal injections in federal death-penalty cases

 As much as I hate capital punishment, but this is good news!

Apparently, the administration of the senile, demented and pathological liar 46th President tried to indirectly abolish the death penalty.

"The Justice Department said many of the limits the Biden administration put on carrying out federal death sentences would be removed. Until recently, firing squads were almost unheard of in modern executions. President Trump, in his first term, reactivated the federal death penalty after a 17-year hiatus and put 13 inmates to death in its final months. Since his return to office, the DOJ is seeking death sentences against 44 defendants."

Wall Street Journal What's news

A majority of health research focuses on men's bodies. Nonprofit FemTechAZ aims to change that. Really!

A case of junk journalism! I have seen at least one other, similar news report this week!

Feminists have been repeating this story since the 1960s or so!

Yes, for historical reasons medicine focused on male bodies in the past, but this is not true anymore for at least the past 50 years. It time to cancel this false, outdated  propaganda!

Caveat: I did not read the entire article.

A majority of health research focuses on men's bodies. Nonprofit FemTechAZ aims to change that

What is wrong with the Israel-Lebanon ceasefire?

A lot! I have been irritated by this so called cease fire agreement since it's inception!

Israel is not at war with Lebanon, but with the Iran sponsored Hezbollah terrorists operating from the territory of Lebanon!

Why was the Lebanese government incapable or unwilling to address the Islamist terrorism attacking Israel on its territory? A huge failure by the Lebanese government! Almost like a failed state!

Google Says 75% of the company's new Code Now Generated by AI

Amazing stuff! Good news! With the help of ML & AI we can now produce so much more programming code (in any programming language)!

"

• Three-quarters of new code at Google is being generated by AI, the company said.
• The number has been steadily increasing as the company pushes staff to adopt AI tools.
• Google CEO Sundar Pichai said a recent code migration was done six times faster thanks to AI agents.

...

 As of October 2024, around a quarter of the company's code was AI-generated, Google said at the time.
Last fall, it said the number had risen to 50%. ..."

Google Says 75% of Fresh Code Now Generated by AI "Google announced this week that 75 percent of all new code created within the company is currently being generated by AI systems and subsequently reviewed by human engineers."

Google says 75% of the company's new code is AI-generated (original news)

Plant seeds can actually sense the sound of rain

Amazing stuff! The sound of rain! Let it rain, let it rain!

"MIT engineers found that plant seeds can actually sense the sound of rain, with rice seeds germinating 30% to 40% faster when exposed to the vibrations of falling water droplets."

"... In experiments with rice seeds, the team found that the sound of falling droplets effectively shook the seeds out of a dormant state, stimulating them to germinate at a faster rate compared with seeds that were not exposed to the same sound vibrations. ..."

"... the first direct evidence that plant seeds and seedlings can sense sounds in nature. Their experiments involved rice seeds that they submerged in shallow water. Rice can germinate in both soil and shallow water. The researchers suspect that many similar seed types may also respond to the sound of rain. ...

They found that when a raindrop hits the surface of a puddle or the ground, it generates a sound wave that makes the surroundings vibrate, including any shallowly submerged seeds. These vibrations can be strong enough to dislodge a seed’s “statoliths,” which are tiny gravity-sensing organelles within certain cells of a seed. When these statoliths are jostled, their movement is a signal for seeds and seedlings to grow and sprout. ..."

From the abstract:

"The ability of natural environmental sound to stimulate seeds and seedlings sufficiently to foster growth has not been previously demonstrated or quantified.

To study this, rain sound is a logical starting point. Rain produces extremely high amplitude sound pressure with commensurate particle displacements in the upper soil, puddles and wetlands where many plant seeds germinate.

Experiments were conducted with controlled rain drops impacting soil and shallow water puddles containing submerged seeds of rice (oryza sativa). Germination rates were measured as the peak sound pressure of drop impact was varied. The displacements of micro-meter-scale statoliths relative to the structure of specialized seed cells that sense gravitational direction were estimated as a function of the controlled rain sound forcing.

The results here indicate rice and related seed types can sense the sound of rain impacting the soil or water surface above them and respond by accelerating germination at depths where impulsive rain sound is sufficiently intense to intermittently shake statoliths from contact with cell membrane receptors and trigger gravitropic growth mechanisms.

The ability to perceive rain sound and respond with accelerated germination is found to be roughly limited to the relatively shallow depths that are also beneficial to seedling survival."

Friday, April 24, 2026 - Join The Flyover

Plants can sense the sound of rain, new study finds

Plants can sense the sound of rain, a new study finds (original news release) "Experiments by MIT engineers show rice seeds sprout faster to the sound of rain."

Seeds accelerate germination at beneficial planting depths by sensing the sound of rain (open access)

Kakapel rock art timeline reveals 9,000 years of painters in Kenya

Amazing stuff! The abstract of this research paper is lousy and dilettante!

"A recent study ... presents the first millimeter-accurate recording of the paintings at Kakapel rock shelter in Kenya, linking the layers of rock art painted over thousands of years to at least three culturally and genetically distinct groups.

The combination of rock art analysis, aDNA, and excavation data makes Kakapel one of the most comprehensively understood rock art sites in eastern Africa. ..."

From the abstract:

"... first described the Kakapel rock art of western Kenya in 1977.  ...

Fieldwork in 2011 was undertaken to ... further research at the site. This paper presents the first millimetre-accurate redrawing of the main panel at Kakapel and places the site within the broader context of eastern African rock art to propose an age and authorship for the different painted traditions represented there.

It thus builds on and extends current understandings of the rock art sequence of eastern Africa."

Kakapel rock art timeline reveals 9,000 years of painters in Kenya

The rock art of Kakapel Shelter, western Kenya (open access)

Figure 4. Kakapel: redrawing of the images in Layer 1 (redrawn by Wendy Voorveld in 2012 and coloured for publication by Kgolagano Vena).

Figure 5. Kakapel: redrawing of the images in Layer 2 (redrawn by Wendy Voorveld in 2012 and coloured for publication by Kgolagano Vena).

Figure 6. Kakapel: redrawing of the images in Layer 3 (redrawn by Wendy Voorveld in 2012 and coloured for publication by Kgolagano Vena).

Figure 1. Map of East Africa showing location of Kakapel rock art site in relation to other sites and places mentioned in the paper.