Evidence worth funding.

Tardigrades are microscopic animals renowned for their ability to survive extreme environmental conditions, including desiccation, freezing, radiation, and the vacuum of space. This study presents the genome of the tardigrade Ramazzottius varieornatus and identifies genetic adaptations associated with extraordinary stress tolerance. The findings provide insight into the molecular basis of extremotolerance and may inform future applications in biotechnology and medicine.

This study reports the discovery of an active microbial ecosystem beneath the West Antarctic Ice Sheet. By analyzing samples from a subglacial lake isolated for thousands of years, researchers identified diverse microorganisms capable of surviving in complete darkness using inorganic compounds as energy sources. The findings provide insights into life in extreme environments and have implications for the search for extraterrestrial life.

This study investigates previously uncharacterized microbial communities inhabiting deep subsurface aquifers. Using metagenomic sequencing, researchers reconstructed genomes from uncultivated microorganisms and identified novel pathways for carbon fixation and energy metabolism. The findings expand current understanding of Earth's deep biosphere and the ecological roles of microorganisms living in nutrient-limited underground environments.

This study presents the genome assembly of the bioluminescent cardinalfish Siphamia tubifer, a reef fish that maintains a highly specific symbiotic relationship with luminous bacteria. The research identifies genomic features associated with the evolution of its light organ and examines adaptations supporting long-term bacterial symbiosis. The findings contribute to understanding host–microbe coevolution in marine ecosystems.

This study examines naturally bioluminescent fungi as ecological indicators of undisturbed tropical forest habitats. Researchers evaluated species distribution, environmental conditions, and seasonal variation to determine whether fungal light emission correlates with habitat quality. Results suggest that glowing fungal communities are closely associated with stable, high-humidity forest environments and may serve as valuable bioindicators for conservation planning.

Wildfires can mobilize heavy metals previously bound within forest soils, creating long-term ecological risks. This research investigates how native mycorrhizal fungi and saprotrophic mycelial networks influence heavy metal immobilization and nutrient cycling during post-fire ecosystem recovery. The findings suggest that fungal communities play a significant role in restoring soil stability while reducing metal bioavailability to regenerating vegetation.

Passive underwater acoustic monitoring has emerged as a low-cost approach for assessing ecosystem health. This study investigates whether acoustic characteristics generated around mangrove root systems correlate with juvenile fish biodiversity. Results demonstrate measurable relationships between soundscape complexity and species richness, suggesting passive acoustic monitoring can supplement conventional biodiversity surveys in sensitive coastal habitats.

This research presents a convolutional neural network (CNN) model for detecting crop diseases from leaf images. The proposed model helps farmers identify diseases early, improving crop yield and reducing losses.

This study proposes a hybrid swarm intelligence and reinforcement learning approach for adaptive urban traffic signal control. The method aims to reduce congestion, improve traffic flow, and optimize signal timing by combining the global search capability of swarm intelligence with the learning ability of reinforcement learning.

In fine art, especially painting, humans have mastered the skill to create unique visual experiences through composing a complex interplay between the content and style of an image. Thus the task of separating and recombining image content and style is of paramount importance to the understanding of human visual perception. Here we introduce an artificial system based on a Deep Neural Network that creates artistic images of high perceptual quality. The system uses neural representations to separate and recombine content and style of arbitrary images, providing a neural algorithm for the creation of artistic images. Moreover, in light of the striking similarities between performance-optimised artificial neural networks and biological vision, our work offers a path forward to an algorithmic understanding of how humans create and perceive artistic imagery.

Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns such as handwritten characters. This paper reviews various methods applied to handwritten character recognition and compares them. Convolutional neural networks are specifically designed to deal with the variability of 2D shapes, and are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation, recognition, and language modeling.

Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly due to insufficient, decaying error backflow. We briefly review Hochreiter's analysis of this problem, then address it by introducing a novel, efficient, gradient based method called Long Short-Term Memory. LSTM can learn to bridge time intervals in excess of 1000 steps even in case of noisy, incompressible input sequences, without loss of short time lag capabilities. This is achieved by an efficient, gradient based algorithm for an architecture enforcing constant error flow through internal states of special units. Key points are constant error carousels and multiplicative gate units that learn to open and close access to the constant error flow.

Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization, and in some cases eliminates the need for Dropout.

Deep neural nets with a large number of parameters are very powerful machine learning systems. However, overfitting is a serious problem in such networks. Large networks are also slow to use, making it difficult to deal with overfitting by combining the predictions of many different large neural nets at test time. Dropout is a technique for addressing this problem. The key idea is to randomly drop units along with their connections from the neural network during training. This prevents units from co-adapting too much. During training, dropout samples from an exponential number of different thinned networks. At test time, it is easy to approximate the effect of averaging the predictions of all these thinned networks by simply using a single unthinned network that has smaller weights.

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small 3x3 convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively.

We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models, BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications.

Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers which is 8x deeper than VGG nets but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task.

We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0% which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully-connected layers we employed a recently-developed regularization method called 'dropout' that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry.

Cytokines are protein mediators of cell-to-cell communication important in a variety of physiologic and pathophysiologic processes. Only a decade ago information about cytokines was limited to the description of a multitude of ill-defined “factors” present in crude super- natants of cultures of mononuclear cells. Modern techniques of protein chemistry and molecular biology have led to the isolation and precise functional definition of many cytokines, including the interferons, colony-stimulating factors, and a heterogeneous group of agents termed interleukins. The name “interleukins” was originally conceived to designate proteins produced by lymphocytes or monocytes, affecting the growth and/or differentiation of lymphoid, monocytic or myeloid cells. With time it has become apparent that many interleukins are not produced exclusively by white blood cells, nor are their actions restricted to leukocytes. The terms “interleukin” or “lymphokine” are now being used for cytokines that can be produced by a variety of cell types and that can affect the functions of many nonhematopoietic cells.

This paper investigates the strategic liquidation and distribution of a finite, state-constrained pool of funding resources within a decentralized science (DeSci) framework. Operating with a rigid ceiling of G=10 available patron grants, we construct an optimal control model using a continuous-time Hamilton-Jacobi-Bellman (HJB) formulation. We derive the exact analytical solution for the optimal payout trajectory under different market regimes of user submission frequency and empirical metadata quality. Finally, we provide a deterministic boundary condition to protect the protocol against treasury exhaustion under sudden surges in data volume.

This paper provides a stochastic analysis of decentralized peer-review systems operating under asymmetric information without cryptographic identity verification. We formalize the "Forum Review" process as a sequential decision-making chain to identify the threshold bounds where rational agents discard private signals to join information cascades (herding). We derive a mathematical proof for an optimal stopping rule that the smart contract layer can deploy to automatically halt consensus voting once a definitive mathematical truth-threshold is reached, thereby minimizing verification latency and exposure to Sybil collusion vectors.

This paper formalizes the cryptoeconomic mechanisms behind the Maecenas Research Funding Protocol. We present an optimization framework for a public source record registry mapping bounded unlock prices in USDC to informational utility curves. We model the protocol's forum review tier as an asymmetric information coordination game, proving that a truth-telling Nash equilibrium is sustained without native wallet identity verification when penalty structures outscale deviation gains. Finally, we establish a finite-horizon stochastic allocation policy for the dynamic depletion of public patron grants.

We present our experience with QUIC, an encrypted, multiplexed, and low-latency transport protocol designed from the ground up to improve transport performance for HTTPS traffic and to enable rapid deployment and continued evolution of transport mechanisms. QUIC has been globally deployed at Google on thousands of servers and is used to serve traffic to a range of clients including a widely-used web browser (Chrome) and a popular mobile video streaming app (YouTube). We estimate that 7% of Internet traffic is now QUIC

This research presents a comprehensive chemical and thermodynamic analysis of volatile compounds embedded within lunar regolith, focusing on their viability for In-Situ Resource Utilization (ISRU). With a specific emphasis on the spatial distribution and concentration of water ice, hydrogen, and carbon oxides across the Moon's permanently shadowed polar regions, this study evaluates extraction efficiencies using advanced thermal desorption methodologies. Experimental and simulated data models demonstrate how optimized heating cycles can maximize yield while minimizing energy expenditure. The final framework provides actionable insights to optimize life support and propellant production capabilities for long-duration lunar habitats, ultimately mitigating the immense logistical costs and risks associated with Earth-reliant replenishment missions.

This paper introduces a groundbreaking approach to the register allocation problem in compilers. By modeling the interference graph of variables (where nodes are variables and edges represent simultaneous live ranges) and applying graph coloring algorithms, Chaitin demonstrates how to efficiently map an unlimited number of program variables to a limited number of physical machine registers. This technique remains a standard pillar of modern optimizing compilers

The dominant sequence transduction models are based on complex recurrent or convolutional neural networks that include an encoder and a decoder. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train.

A new method for determining nucleotide sequences in DNA is described. It is similar to the "plus and minus" method but makes use of the 2',3'-dideoxy and arabinonucleoside analogues of the normal deoxynucleoside triphosphates, which act as specific chain-terminating inhibitors of DNA polymerase.

We wish to suggest a structure for the salt of deoxyribose nucleic acid (D.N.A.). This structure has novel features which are of considerable biological interest. A structure for nucleic acid has already been proposed by Pauling and Corey... We wish to put forward a radically different structure for the salt of deoxyribose nucleic acid. This structure has two helical chains each coiled round the same axis.

Determinations of the distances of extra-galactic nebulae from Cepheid variables are combined with their radial velocities to demonstrate a linear relationship between distance and velocity. This fundamental observation establishes the expansion of the universe.

The presence of a Jupiter-mass companion to the star 51 Pegasi is inferred from observations of periodic variations in the star's radial velocity. The companion lies only about eight million kilometres from the star, which would be well inside the orbit of Mercury in our Solar System.

We present spectral and photometric observations of 10 type Ia supernovae (SNe Ia) in the redshift range 0.16 < z < 0.62. The data indicate that the cosmological expansion is accelerating, and imply the existence of a cosmological constant or similar "dark energy".

On September 14, 2015, the two detectors of the Laser Interferometer Gravitational-Wave Observatory simultaneously observed a transient gravitational-wave signal. The signal sweeps upwards in frequency and matches the waveform predicted by general relativity for the inspiral and merger of a pair of black holes and the ringdown of the resulting single black hole.

Artificial intelligence (AI) and neuroscience have a long history of working together. This research explores whether the human brain and deep neural network (DNN) models recognize objects using comparable steps of computation. By analyzing the hierarchical, brain-inspired organization of computations within machine learning systems, this study investigates the overlap between biological processes and artificial intelligence to improve real-world applications such as facial recognition and healthcare models.

This study highlights the influence of recent developments in compiler design and optimization techniques on program execution speed, memory utilization, and overall software quality. It covers language-specific optimization techniques such as Just-In-Time (JIT) compilation and Profile-Guided Optimization (PGO), loop optimization techniques like loop unrolling and loop fusion, as well as automatic parallelization, SIMD vectorization, and memory hierarchy optimizations.

SUMMARY The responses of microorganisms (viruses, bacterial cells, bacterial and fungal spores, and lichens) to selected factors of space (microgravity, galactic cosmic radiation, solar UV radiation, and space vacuum) were determined in space and laboratory simulation experiments. In general, microorganisms tend to thrive in the space flight environment in terms of enhanced growth parameters and a demonstrated ability to proliferate in the presence of normally inhibitory levels of antibiotics. The mechanisms responsible for the observed biological responses, however, are not yet fully understood. A hypothesized interaction of microgravity with radiation-induced DNA repair processes was experimentally refuted. The survival of microorganisms in outer space was investigated to tackle questions on the upper boundary of the biosphere and on the likelihood of interplanetary transport of microorganisms. It was found that extraterrestrial solar UV radiation was the most deleterious factor of space. Among all organisms tested, only lichens (Rhizocarpon geographicum and Xanthoria elegans) maintained full viability after 2 weeks in outer space, whereas all other test systems were inactivated by orders of magnitude. Using optical filters and spores of Bacillus subtilis as a biological UV dosimeter, it was found that the current ozone layer reduces the biological effectiveness of solar UV by 3 orders of magnitude. If shielded against solar UV, spores of B. subtilis were capable of surviving in space for up to 6 years, especially if embedded in clay or meteorite powder (artificial meteorites). The data support the likelihood of interplanetary transfer of microorganisms within meteorites, the so-called lithopanspermia hypothesis.

SUMMARY Biology today is at a crossroads. The molecular paradigm, which so successfully guided the discipline throughout most of the 20th century, is no longer a reliable guide. Its vision of biology now realized, the molecular paradigm has run its course. Biology, therefore, has a choice to make, between the comfortable path of continuing to follow molecular biology's lead or the more invigorating one of seeking a new and inspiring vision of the living world, one that addresses the major problems in biology that 20th century biology, molecular biology, could not handle and, so, avoided. The former course, though highly productive, is certain to turn biology into an engineering discipline. The latter holds the promise of making biology an even more fundamental science, one that, along with physics, probes and defines the nature of reality. This is a choice between a biology that solely does society's bidding and a biology that is society's teacher.

Aging is characterized by an extensive remodeling of epigenetic patterns, which has been implicated in the physiopathology of age-related diseases. Nutrition plays a significant role in modulating the epigenome, and a growing amount of data indicate that dietary changes can modify the epigenetic marks associated with aging. In this review, we will assess the current advances in the relationship between caloric restriction, a proven anti-aging intervention, and epigenetic signatures of aging. We will specifically discuss the impact of caloric restriction on epigenetic regulation and how some of the favorable effects of caloric restriction on lifespan and healthspan could be mediated by epigenetic modifications.

Clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated (Cas) systems provide bacteria and archaea with adaptive immunity against viruses and plasmids by using CRISPR RNAs (crRNAs) to guide the silencing of invading nucleic acids. We show here that in a subset of these systems, the mature crRNA that is base-paired to trans-activating crRNA (tracrRNA) forms a two-RNA structure that directs the CRISPR-associated protein Cas9 to introduce double-stranded (ds) breaks in target DNA. At sites complementary to the crRNA-guide sequence, the Cas9 HNH nuclease domain cleaves the complementary strand, whereas the Cas9 RuvC-like domain cleaves the noncomplementary strand. The dual-tracrRNA:crRNA, when engineered as a single RNA chimera, also directs sequence-specific Cas9 dsDNA cleavage. Our study reveals a family of endonucleases that use dual-RNAs for site-specific DNA cleavage and highlights the potential to exploit the system for RNA-programmable genome editing.

Inhibition of the TOR signalling pathway by genetic or pharmacological intervention extends lifespan in invertebrates, including yeast, nematodes and fruitflies; however, whether inhibition of mTOR signalling can extend lifespan in a mammalian species was unknown. Here we report that rapamycin, an inhibitor of the mTOR pathway, extends median and maximal lifespan of both male and female mice when fed beginning at 600 days of age. On the basis of age at 90% mortality, rapamycin led to an increase of 14% for females and 9% for males. The effect was seen at three independent test sites in genetically heterogeneous mice, chosen to avoid genotype-specific effects on disease susceptibility. Disease patterns of rapamycin-treated mice did not differ from those of control mice. In a separate study, rapamycin fed to mice beginning at 270 days of age also increased survival in both males and females, based on an interim analysis conducted near the median survival point. Rapamycin may extend lifespan by postponing death from cancer, by retarding mechanisms of ageing, or both. To our knowledge, these are the first results to demonstrate a role for mTOR signalling in the regulation of mammalian lifespan, as well as pharmacological extension of lifespan in both genders. These findings have implications for further development of interventions targeting mTOR for the treatment and prevention of age-related diseases.

This foundational paper introduces the concept of the epigenetic clock, demonstrating that DNA methylation levels across various CpG sites can accurately predict the biological age of human tissues and cell types. This method outperforms traditional chronological age markers and provides a unified biological age estimator for cellular longevity studies.

Differentiated cells can be reprogrammed to an embryonic-like state by transfer of nuclear contents into oocytes or by fusion with embryonic stem (ES) cells. Little is known about factors that induce this reprogramming. Here, we demonstrate induction of pluripotent stem cells from mouse embryonic or adult fibroblasts by introducing four factors, Oct3/4, Sox2, c-Myc, and Klf4, under ES cell culture conditions. Unexpectedly, Nanog was dispensable. These cells, which we designated iPS (induced pluripotent stem) cells, exhibit the morphology and growth properties of ES cells and express ES cell marker genes. Subcutaneous transplantation of iPS cells into nude mice resulted in tumors containing a variety of tissues from all three germ layers. Following injection into blastocysts, iPS cells contributed to mouse embryonic development. These data demonstrate that pluripotent stem cells can be directly generated from fibroblast cultures by the addition of only a few defined factors.

Physical function declines in old age, portending disability, increased health expenditures, and mortality. Cellular senescence, leading to tissue dysfunction, may contribute to these consequences of aging, but whether senescence can directly drive age-related pathology and be therapeutically targeted is still unclear. Here we demonstrate that transplanting relatively small numbers of senescent cells into young mice is sufficient to cause persistent physical dysfunction, as well as to spread cellular senescence to host tissues. Transplanting even fewer senescent cells had the same effect in older recipients and was accompanied by reduced survival, indicating the potency of senescent cells in shortening health- and lifespan. The senolytic cocktail, dasatinib plus quercetin, which causes selective elimination of senescent cells, decreased the number of naturally occurring senescent cells and their secretion of frailty-related proinflammatory cytokines in explants of human adipose tissue. Moreover, intermittent oral administration of senolytics to both senescent cell-transplanted young mice and naturally aged mice alleviated physical dysfunction and increased post-treatment survival by 36% while reducing mortality hazard to 65%. Our study provides proof-of-concept evidence that senescent cells can cause physical dysfunction and decreased survival even in young mice, while senolytics can enhance remaining health- and lifespan in old mice.

This paper explores how decentralized systems and DAOs provide a new paradigm for scientific development. It addresses how tokenized structures can solve information silos and funding bottlenecks in traditional science by decentralizing decision-making authority away from central gatekeepers.

Rapid advancements in Distributed Ledger Technology (DLT), including blockchain, are foundational to a new era of digital innovation. This innovation has catalyzed the emergence of 'Decentralized Science (DeSci)'. DeSci aims to address pressing challenges in modern science, such as the scientific publishing system, insufficient diversity, and a lack of transparency in funding allocation. This study provides a comprehensive definition of the term and explores shared values and guiding principles inherent to DeSci.

Aging is driven by hallmarks that encompass biochemical changes occurring in all organisms that experience aging, leading to a progressive loss of physiological integrity, impaired function, and increased vulnerability to death. This review enumerates the core hallmarks that represent common denominators of aging, with a special emphasis on mammalian aging. These hallmarks include: genomic instability, telomere attrition, epigenetic alterations, loss of proteostasis, deregulated nutrient sensing, mitochondrial dysfunction, cellular senescence, stem cell exhaustion, and altered intercellular communication.

Web3 today centers around expressing transferable, financialized assets, rather than encoding social relationships of trust. Yet many core economic activities—such as uncollateralized lending and building personal brands—are built on persistent, non-transferable relationships. In this paper, we illustrate how non-transferable "Soulbound" tokens (SBTs) representing the commitments, credentials, and affiliations of "Souls" can encode the trust networks of the real economy to establish provenance and reputation. We highlight applications of SBTs to sybil-resistant governance, decentralizing market mechanisms, and novel markets with decomposable, shared rights.

Coinbase is working with AWS CloudFront and WAF to let the web publishers and API providers behind a quarter of the internet accept agents as customers via x402

Compares subscriptions, card fees, account minimums, and settlement delays against sub-cent software payments.

A critique of pay-per-citation markets, including spam incentives, ownership ambiguity, and quality risks.

A short research memo on why generated answers need explicit evidence trails and source-level accounting.

A security-oriented overview of agent wallets, spending limits, and budget policies for autonomous software.

A creator economy essay arguing that AI usage should produce small direct rewards for the sources that shape generated work.

Notes on stablecoin settlement assumptions for low-value machine payments on Arc-compatible infrastructure.

Explains how HTTP 402 can become an agent-readable negotiation point for paid API access and source evidence.

A technical note on how Gateway-style nanopayments let software agents purchase API calls and evidence fragments without account contracts.

A broad overview of blockchain payments with limited detail on AI agents, evidence pricing, or x402 workflows.

A product architecture note for APIs that quote prices dynamically to machine clients.