Harvard/Zitnik Lab presented ATHENA-R1: an AI agent selects treatments from FDA-approved drugs and explains its decision using verifiable sources.

The system operates as a search chain: it decides what data is needed, invokes biomedical tools, and compiles a response with a visible evidence trail. The authors tested it on drug problems, patient scenarios, expert assessments of rare diseases, and historical data from 5.4 million patients.

In medicine, choosing the right drug from existing options is often difficult. The patient has a disease, age, kidneys, liver, other medications, contraindications, risk of side effects, and incomplete recommendations. In such a task, confident text is dangerous: the doctor needs to see the origin of each step.

On June 27, the team of Shanghua Gao and Marinka Zitnik from Harvard Medical School published the ATHENA-R1 preprint. The model, based on Qwen3-8B, was trained to work with 212 biomedical tools. These tools access open sources like the FDA's drug label database, Open Targets, and Human Phenotype Ontology, an ontology that links human signs and symptoms to diseases.

The project's website provides a simple example: an elderly patient with diabetes, hypertension, and early chronic kidney disease is taking metformin. ATHENA-R1 must check dosages for reduced kidney function, interactions with other medications, warnings from the label, and suitable alternatives. Finally, it provides a recommendation and a reasoning trail: which sources it accessed and what it learned from each.

The authors call this treatment reasoning. The Russian translation is that the system selects a therapy step by step based on the patient's limitations. This is a familiar problem, amplified for future anti-aging medicine: geroscience drugs, senolytics, mTOR modulators, GLP-1 agents, anti-inflammatory regimens, and cell therapies will all face comorbidities, drug interactions, and weak endpoints.

A few days ago, MIRA and AMIE tested medical AI agents in a virtual clinic: one agent worked in an electronic health record sandbox, while the other guided an actor patient through three outpatient visits. ATHENA-R1 takes the next step in the same story: drug selection, dosage, and constraints from external sources, followed by a visible trace of how the model arrived at its answer.

In the preprint, ATHENA-R1 scored 94.7% on 3,168 drug-data tasks and 82.9% on 456 patient-specific scenarios. GPT-5 scored 76.9% and 72.2% in the same open evaluations. To reduce the risk of memorization, the authors based some of the tests on FDA-approved drugs approved in 2024, and excluded drugs approved after 2023 from the training.

The team recruited experts through 28 rare disease organizations; Twenty-three raters blindly compared 110 ATHENA-R1 responses with responses from other models and favored ATHENA-R1 more often across eight criteria, including accuracy, clinical relevance, and clarity of the chain. The team then took ATHENA-R1's adverse event hypotheses and tested them on Clalit Health Services' electronic medical data: three of the six predictions yielded statistically significant increases in risk in the relevant patient groups.

The system's status is limited. The project's GitHub page describes ATHENA-R1 as a research artifact for studying treatment reasoning and decision support; it has not been approved for clinical use. Retrospective validation reveals associations in past data, and the benefit of prescribing treatment based on model advice should be verified by future studies.

Medical AI is gradually moving away from "memory-based" responses to a procedure in which the model must know what to look for, where to check, and how to show a trace. For longevity, such a procedure is more beneficial than yet another confident assistant: the fight against aging will rest on the ability to safely select interventions for living, complex individuals already undergoing treatment.

A portion of the tech elite is starting to serve up a selection of embryos and future editing of heredity as a way to catch up with superintelligence. On April 16, Mother Jones published a large exposé on how former MIRI researcher Tsvi Benson-Tilsen, the Berkeley Genomics project, and a circle of related investors are linking the fear of AGI (Artificial General Intelligence) with the market for "enhanced children." Here, embryo selection, discussions of future heredity editing, and the money of people invested in both AI and genomics startups converge. AGI refers to an artificial general intelligence system that can solve a wide range of tasks at or beyond the level of human capabilities. In the AI-risk environment, the fear is formulated as follows: the next step for such a system may be to become superhuman, and humans will no longer understand its goals and lose control. Benson-Tilsen worked for seven years at the Machine Intelligence Research Institute, where they tried to solve this very problem. According to Mother Jones, he came to the conclusion that he himself was not capable of solving it, and that others had not succeeded either. His response now is biological. At the end of 2024, he launched the Berkeley Genomics Project. On the project's website, the mission is stated directly: to open the path to safe and accessible heritable genome engineering, i.e., to modify DNA in an embryo or reproductive cells so that these edits are inherited not only by the child but also by their descendants. Among the promised benefits listed are protection from diseases, protection from severe mental disorders, a longer life and more healthy years, and a "more capable mind." The political continuation is also stated: the USA should lead in this technology. It is essential to distinguish between two things. Today, the market primarily sells embryo selection: during IVF, several embryos are obtained, their DNA is checked, and parents are helped to choose one for transfer. For some severe hereditary diseases, such a check is understandable. For complex traits like intelligence, everything is much weaker. They depend on many genes, as well as the environment, nutrition, family, and school. A calculation in Cell in 2019 gave an average expected gain of about 2.5 IQ points when choosing from five embryos. This is little. The history of SAT scores has shown how quickly the conversation about "innate abilities" begins to confuse heredity with language, school, and environment. Until the full-fledged "construction of geniuses," the market has not yet reached. But the language of the field has already moved further than practice. Brian Armstrong described a future IVF clinic in his post on X as a "Gattaca screen": "The IVF clinic of the future will combine several technologies: the production of egg cells from skin or blood, the selection of an embryo that best fits the parents' request, ideally from thousands of options, editing the embryo for disease prevention or improvement, and artificial wombs." In this formula, an entire conveyor belt is assembled. First, egg cells are made from skin or blood cells. Then, many embryos are created. Then, they are compared according to genetic probabilities, one is chosen, additional edits are made, and the pregnancy is carried out in an artificial womb. Moreover, the artificial womb is already being assembled: there are models of embryos, artificial placentas, and a hermetically sealed environment for the development of the fetus outside the body. A set of controversial technologies is already being presented as a single product roadmap. The topic goes beyond the debate about fertility and disease prevention. The same tech elite networks are simultaneously fueling the AGI agenda, financing AI, and starting to justify heritable human upgrades as a response to the risk from these systems. Yesterday, the market was selling wealthy parents a slightly more advantageous choice between embryos — we already had a sepa…

🔗 Read original →

A major review: weight loss, heart protection, and long life in obesity medications differ. On July 8, BMJ published a review of 262 randomized trials - studies where participants are randomly assigned a medication or a comparator - with nearly 100,000 participants. The medications differ significantly in terms of weight, cardiac outcomes, side effects, and quality of life. Subcutaneous semaglutide was the only medication with a convincing reduction in overall mortality; these data came mainly from studies of people with already high cardiovascular risk. The market has become accustomed to comparing such medications by the number of kilograms lost. For someone who wants to live longer, this scale is too crude. Weight may decrease quickly, but the data on survival, heart function, and well-being remain different for each medication. The review authors compared 19 medications and found: in terms of weight loss, tirzepatide and the combination of cagrilintide with semaglutide led. Subcutaneous semaglutide reduced body mass by approximately 9.8% more strongly than a single lifestyle change. But among all medications, only it showed a convincing reduction in overall mortality: by 19% in 17 trials with 25,264 participants. The risk of myocardial infarction was 28% lower. The reduction in mortality in this review refers to subcutaneous semaglutide in people with existing cardiovascular diseases. For people without such diseases and for medications with similar effects, separate data are needed: absolute risks and outcomes may differ. Tirzepatide, for example, reduced the risk of heart failure, and the data on its effect on overall mortality have lower confidence. Weight loss and quality of life assessments differed. The authors considered a clinically significant difference to be 10 points on the well-being scale. All medications remained below this threshold; subcutaneous semaglutide had a score of 2.9 points. In the study of body composition with fat, tirzepatide also reduced lean mass - tissues that include muscles, water, and other components of the body. For each medication, one must ask which specific outcome it changes: weight, myocardial infarction, mortality, ability to move, or everyday well-being. Then - in whom exactly was this tested and how long was it observed. The scale on the weights answers its own question; prolonging life requires data on mortality and organ function. Source: BMJ Telegraph.

🔗 Read original →

Professor of Yale Samuel Moyn suggests giving the young more votes: the debate on aging has already become a debate on rights. In July interviews about the book "Gerontocracy in America", Moyn asserts that power, money, and voters in the US are too skewed towards older generations. Among his ideas are age limits for part of the positions and different political weight for people with different remaining life expectancy. On July 13, The New York Times made a separate video report on this debate. Yale law and history professor Samuel Moyn has just released the book "Gerontocracy in America", around which the discussion has unfolded. Gerontocracy, in his definition, is the concentration of power among older generations through money, property, and the setup of elections. In a July 9 interview with Yale University, Moyn cites such figures: the age of half of the members of the US Congress is over 60, half of the voters are 52 and older, and workers over 55 make up about a quarter of the workforce, compared to 10% in 1990. He links this picture to the fact that it is harder for young people to get housing, jobs, and political representation. Moyn does not limit himself to changing faces in Congress. He suggests discussing age limits for part of the positions, compulsory voting, and a different way of counting votes. In a debate with political scientist Yascha Mounk, he explains his logic as follows: a person who will live longer with the consequences of a law has a greater stake in it. Hence the idea of giving young people more political weight. Here, the debate on aging turns into a debate on the value of remaining life. Moyn wants to protect future generations from decisions that they will live with longer than everyone else. But age poorly separates power from powerlessness. He himself acknowledges that elderly people without means also suffer from the current system. Wealth and influence are concentrated among a minority, and age often coincides with this concentration. Radical life extension makes his argument more dangerous and clearer. If political weight depends on life expectancy, the state gets a reason to reduce a person's vote precisely because they will live longer. Health and survival become the basis for reducing political rights. The same problem has already arisen in the debate on space resources, which can determine human lifespan. There, the rich can buy more time to live. With Moyn, lifespan becomes the criterion by which the state gives different political weight. In both cases, access to rights depends on the length of life. Gerontocracy arises where power is not changing, wealth is accumulating, and access to decisions is closed. Young people need representation and protection of the future; older people need full rights and reliable social protection. The struggle against death requires a common right to a long life, rather than the distribution of civil rights by age.

🔗 Read original →

#digest for July 14 Full article • The Alzheimer’s Association will invest $100 million to investigate: does the medication add protection against dementia to the prevention program • At the ICML workshop, Bruno was presented - an AI assistant that is supposed to keep track of the work of a scientific group • Takeshi Kozai: a neurointerface in 20 years should preserve living tissue around the electrode • JAMA: in Americans of the same age, dementia over 40 years has become approximately two-thirds less common • A large review: weight loss, heart protection, and long life in obesity medications diverge • Some techno-elite are starting to offer embryo selection and future gene editing as a way to catch up with superintelligence • Professor Samuel Moyn of Yale proposes giving the young more votes: the debate about aging has already become a debate about rights Telegraph Digest @UkhvatNews - July 14 When talking about the "wave of dementia", people usually combine two different facts into one sentence. The first fact: there are more people living up to 80, 90, and 100 years. The second: the likelihood of losing memory, orientation, and independence increases with age. From this, it is easy…

🔗 Read original →

Demis Hassabis proposes creating a US agency to test advanced AI systems before release. The head of Google DeepMind suggests creating a US organization that would determine the threshold for advanced AI and test such models before release. The initial tests would be voluntary; if the methodology proves effective and reliable, Hassabis proposes making passing the test a condition for a model to operate on the American market. On July 14, Hassabis published an essay on rules for advanced AI systems. In his scheme, a model is considered "advanced" if it achieves a set result in test tasks. The new organization would set this result and regularly change the tasks as the capabilities of the models grow. This would determine which laboratories would be subject to a special testing regime. Hassabis takes FINRA, a private US organization that oversees brokerage firms under the supervision of the US Securities and Exchange Commission, as a model. According to his idea, the council of the new agency should include independent technical specialists, as well as representatives of the state, industry, and open-source software developers. "Initially, laboratories would voluntarily submit models to the agency for testing no later than 30 days before release." The tests should assess cyber risks, biological threats, and other high-risk areas; Hassabis separately considers nuclear risks as potential threats. Separate tests could look for attempts to bypass built-in restrictions or signs of deception. If the methodology proves effective and reliable, Hassabis proposes making passing the test a condition for releasing an advanced model in the US. CAISI, a center at the US National Institute of Standards and Technology, has already gained access to Google DeepMind models before public release to assess bio-risks, cyber threats, and risks to critical infrastructure. In Hassabis' proposal, the new agency would itself establish the threshold beyond which a model would be subject to testing. If testing becomes mandatory, the agency would determine the threshold for an "advanced" model and the set of tasks it must pass to enter the US market. Hassabis allows for the possibility of coordinating the slowing down of development between laboratories with models of the advanced class if the situation requires it. Large-scale testing requires computational power and specialists, so Hassabis expects funding from the industry. The first methodologies would be developed in consultation with advanced-class laboratories. Then, the agency should build the capacity to create independent tests, with which developers are not familiar in advance. Such tasks reduce the chance of tailoring a model to known questions. The budget, council composition, and ability to create its own tests would determine the independence of the future agency. Under universal AI, Hassabis understands a system with a human set of cognitive abilities and expects it to accelerate science, medicine, and drug discovery. His proposal provides for testing advanced models before releasing them to the US market.

🔗 Read original →

In long-lived mammals, the SIRT6 protein has more sites for phosphate marks than in short-lived species. The group of Vera Gorbunova and Andrei Seluanov linked this feature to a stronger interaction between SIRT6 - a protein involved in DNA repair - and PARP1. PARP1 recognizes DNA damage and assembles repair proteins. In human cells, mimicking one phosphate mark helped cells survive oxidative damage. In an article published on July 8, researchers studied the flexible C-terminus of SIRT6. Phosphorylation is the addition of a phosphate group to a protein, which changes the protein's electric charge and its interaction with other proteins. The authors compared the SIRT6 sequence of more than 150 mammals with their maximum lifespan. In longer-lived species, this tail had more potential sites for phosphate marks. One of these sites, T294, is present in humans and absent in mice. Using the CRISPR method of precise DNA editing, researchers replaced T294 in human fibroblasts, cells of connective tissue. The T294E variant mimics phosphorylation: after treatment with hydrogen peroxide, such cells survived better. The T294A variant excludes phosphorylation at this position; SIRT6 with this substitution bound weaker to PARP1. The comparison of species linked the number of potential phosphorylation sites to maximum lifespan. The authors counted these sites in sequences; they measured the amount of phosphorylated SIRT6 only in cells of several species. After simultaneously accounting for body mass and species relatedness, the connection between T294 and longevity became insufficiently convincing for a confident conclusion. Therefore, further verification requires mice with a point substitution in the Sirt6 gene and simultaneous measurement of health and lifespan. In a February preprint from the same laboratory, researchers introduced in mice a constant mimicry of phosphorylation at another SIRT6 site, S10E. After irradiation, DNA repair in these mice was better. The median lifespan of males decreased by 10%. After irradiation, in the blood cells of S10E mice, LINE1 expression was higher; in the intestine and brain of individual groups, the authors saw a tendency towards its increase. LINE1 is mobile repetitive fragments of the genome: they create new DNA copies, and accumulated LINE1 DNA in the cytoplasm triggers an inflammatory response. The authors suggest that weakening this control shortened the life of S10E males. For T294, the experiment on mice should immediately measure DNA repair, LINE1 suppression, inflammation, and lifespan. Such a set of results will show whether enhanced DNA repair maintains protection from LINE1 and whether T294 changes lifespan.

🔗 Read original →

Chai Discovery has raised $400 million: AI-designed molecules enter research programs at Pfizer and Novartis. On July 14, Chai Discovery announced a $400 million Series C round at a valuation of $3.8 billion. The funds will go towards computations, data, research, and product development; previously, Pfizer and Novartis had signed separate agreements with the company. The search for a new protein drug begins with a question: which molecule is worth synthesizing and testing in the first place? An antibody is a protein that must recognize and bind to a specific biological target. There are too many amino acid sequence variants to blindly test them in a laboratory. Chai builds models that predict molecular interactions and propose protein candidates with specified properties. These models help select variants for initial laboratory experiments; then, candidates undergo target binding testing, animal testing, and human trials. In June, Biohub models had already yielded binding proteins that had reached laboratory testing. Chai sells pharmaceutical companies the same early stage of molecule search as a working tool. On June 5, Pfizer signed a licensing agreement with Chai. The company will receive early access to Chai-3 and a separate model that uses Pfizer's closed data and is tailored to its own way of searching for drugs. Pfizer is connecting molecular AI to its closed data and development workflows. On July 13, the day before the round announcement, Novartis announced a collaboration with Chai to search for therapeutic antibodies for several targets at once. Prior to this, the companies had been working together technically for over a year; now, Novartis is gaining access to Chai-3 for its own therapeutic programs. The new Chai round will go towards computations, data, research, and product development. The company plans to expand its computational capabilities and the data on which it builds new model versions. The first test of these deals will be the candidates that pass laboratory testing and enter clinical programs.

🔗 Read original →

Limiting calories causes pancreatic cells to conserve calcium instead of a large overall insulin release. In mice that received 20% less food for eight weeks, pancreatic cells changed the rhythm of calcium impulses and withstood artificially induced stress longer. The work shows how a lower demand for insulin by the body can unload a specific type of cell. On June 18, in Cell Calcium, a team from Vienna and Vanderbilt University published a study on beta cells - pancreatic cells that release insulin. The authors sought to answer a simple question: how does a lower demand for insulin by the body change the functioning of these cells? A beta cell releases insulin after a calcium surge. Part of the calcium is stored in the endoplasmic reticulum - an internal reservoir of the cell that also helps to collect proteins. If the reservoir empties quickly, the cell loses its stable rhythm of operation. The authors fed young male mice 20% less food than usual for eight weeks, then observed calcium signals in fresh pancreatic sections. In animals with calorie restriction, beta cells sent shorter and more frequent impulses. Within the islet of Langerhans - a cluster of beta cells that usually release insulin together - their signals coincided in time less. For a healthy body, this is similar to a change in mode. Tissues became more sensitive to insulin and required less of it. Beta cells did not have to gather in a dense network for a simultaneous large release of the hormone. They could work more separately and expend calcium in small portions. Then, the researchers gave the cells a high dose of acetylcholine. It opens the IP3R channel in the endoplasmic reticulum and forces it to quickly release calcium. In control mice, the fluctuations soon subsided; in mice with calorie restriction, they continued longer. Their calcium reserve withstood the load better. This continues the work of the same group from 2024: then, it showed that calorie restriction in mice increases insulin sensitivity, maintains the state of beta cells, and reduces their turnover. The new article adds a possible mechanism: lower external demand changes the rhythm of calcium, cell synchrony, and the ability to withstand a sharp expenditure of the internal reserve. Synchrony depends on the state of the animal. In 2020, another group saw how calorie restriction restored the coordinated work of beta cells in prediabetic mice with obesity. Healthy mice in the new study required less insulin, so the signals within their islets became more separate. In these two models, beta cells adjusted their overall rhythm to the body's demand for insulin. The authors studied eight weeks of feeding in young male mice and acute stress in pancreatic sections. The lifespan of the animals, the risk of diabetes, and the work of beta cells in humans were not included in this study. In the CALERIE human trial, other effects of moderate calorie restriction have already been seen, such as a decrease in C3a - an inflammatory signal of the immune system. The study does not report on the calcium work of the pancreas: we analyzed it separately. In this model, protection arises before damage: tissues require less insulin, so beta cells less often expend the entire calcium reserve on peak loads.

🔗 Read original →

Anil Seth: the working space found in Claude shows reasoning, but does not prove experience. On July 15, a professor of cognitive and computational neuroscience responded to Anthropic's work, which found an internal common channel for intermediate thoughts in Claude. Seth acknowledges the result itself, but believes that the similarity to one of the theories of consciousness does not allow us to conclude that the language model feels anything. On July 6, Anthropic described a narrow section of computations within Claude, which it called J-space. The model holds a word or intermediate result there and then uses it in different tasks. Researchers replaced the internal representation of "spider" with "ant" - and Claude instead of eight named six legs. When they suppressed J-space, the model retained fluent speech and simple responses, but almost lost multi-step reasoning. Such a section is similar to what the theory of global workspace is looking for in the brain: information that becomes available to many systems at once. In everyday life, a person can hold a number, intention, or image in mind, tell about it, and apply it in a new action. Anthropic showed a functional analogue of this operation in Claude: a common internal representation can be read, modified, and causally linked to the model's response. Already on June 2, Google DeepMind, Anthropic, and Meta had referred the possible experience of models to a research task: DeepMind hired a philosopher, and Anthropic is leading a program to study model welfare. J-space gives this debate a measurable subject - an internal representation, the role of which can be changed in experience. In a column for The Guardian, Anil Seth draws a line between such access to information and subjective experience. The first answers the question of whether a system can hold a thought, report it, and use it for selection. The second is whether this system has its own "what it feels like": pain, color, fear, pleasure. Seth points to recurrent loops - feedback cycles in which signals repeatedly pass through brain circuits and support representation over time. In Claude, J-space processing fits into one pass through the network; Anthropic itself notes this distinction. Seth links consciousness to how a living nervous system works together with the body and the world, and not just with the calculations of a silicon program. Seth believes that similar computational organization is not enough to conclude experience. Anthropic limits its result to functional "conscious access": the model can hold content, report it, and use it in reasoning. The authors do not attribute subjective experience to Claude. The future transfer of personality will have to be tested along two different lines. The system must retain memory, goals, and accessible reasoning to continue acting like the original person. Separately, the theory of consciousness must explain what properties of the carrier support its internal experience. J-space provides a way to measure the first line in a language model; the second remains a subject of debate about the brain, body, and consciousness.

🔗 Read original →

AI 2027 underestimates synthetic biology: it may launch a self-replicating economy earlier than robotized factories. On July 14, an author from LessWrong proposed considering an earlier transition to the smallest systems capable of copying themselves from available raw materials. One of the co-authors of the scenario supported part of this criticism. The AI 2027 already has a distant variant with biological infrastructure. In the AI 2027 scenario, the US and China create special economic zones with simplified construction rules. In the American zones, Agent-5 - a superintelligence from the scenario - designs robots and, through humans, manages the construction of factories and laboratories; new factories produce even more robots. As a reference, the authors cite a car factory that in less than a year produces a mass of cars comparable to its own. Then they suggest that an autonomous robotic economy could reproduce itself faster than a year. In response to this scenario, Thomas B. suggests looking for the future leap below the factory scale. He suggests that it would be beneficial for the superintelligence to reduce the unit of self-replication to the limit and design organisms and nanotechnology systems that grow and copy themselves using the surrounding raw materials. The argument is based on the speed of experimentation. In Thomas B.'s opinion, the AI 2027 assumption about robots at the human level would mean automated laboratories without a shortage of qualified human labor. He suggests the next step: such laboratories could create many cheap living sensors and executive systems, conduct experiments with them in parallel, and give the results to simulations. Improved simulations help design the next version of organisms and devices. This research reduces the time for the next research. The dispute comes down to the speed of feedback. A factory needs buildings, machines, supplies, and large machines. A self-replicating biological system could become a sensor, an object of parallel experiments, and part of the next production system. Thomas B. suggests putting this possibility at the center of the scenario. AI 2027 itself is already moving in this direction. The authors cite plants, insects, and bacteria as benchmarks for reproduction rates. In a distant variant, they suggest an economy similar to an ecosystem of new algae: some organisms grow, others process them into materials for floating factories. The disagreement concerns the time of emergence of this biology. The main line first unfolds a robotic economy; Thomas B. considers biological self-replicating systems a likely early driver of acceleration. Thomas Larsen, one of the co-authors of AI 2027, in the comments, agreed with the main criticism and wrote that the finale was worth linking more strongly to nanotechnology. The dispute shifts the question about superintelligence from the number of robots to the length of the "experiment - data - new system" cycle: if this cycle can be compressed to the speed of reproduction of created organisms, factories will cease to be the only measure of production capacity.

🔗 Read original →

Dextran, albumin, and sialic acid helped preserve extracellular vesicles for a year at -20 °C. Chinese researchers have assembled a cryoprotective mixture, DDAS, for extracellular vesicles - tiny membrane particles that cells use to transport proteins and nucleic acids. After storage at -20 °C and repeated thawing, these therapy candidates retained more structure and biological activity than in a conventional buffer. Extracellular vesicles are being considered as a future means of delivering stem cell signals, proteins, and RNA to the body. Such a preparation must travel from the laboratory to the patient, survive storage, and remain the same preparation. When frozen, vesicles aggregate, their membranes rupture, and their contents leak out. Their action is determined by the molecular cargo: four microRNAs in vesicles from an aging liver enhanced metastases in mice. Therefore, for therapy, it is not enough to preserve the number of particles; it is necessary to preserve the specified composition and function. A conventional phosphate buffer poorly solves this problem. DMSO and glycerin can protect biomaterial, but they cannot be left in the preparation for direct intravenous injection: after thawing, additional purification is required. The same engineering problem already exists for platelets: in a laboratory test, a low-dose DMSO regimen without washing returned 94.4% of cells after thawing. The authors of DDAS checked whether it is possible to store vesicles so that after thawing, they can be administered without additional purification. On July 13, an early unedited version of the work on DDAS was published in the Journal of Nanobiotechnology. The mixture includes dextran, albumin, and sialic acid; the authors selected it by starting with a list of molecules found in human body fluids. Generative models narrowed down the list of candidates, and the composition was then tested on vesicles, cells, and mice. Vesicles from brain vascular endothelial cells were stored at -20 °C for a year. DDAS better preserved the number of particles, RNA, and membranes than a conventional buffer; by the 12th month, the authors recovered around 50% of nucleic acids. After five cycles of freezing and thawing, it also outperformed the buffer and a commercial cryoprotector. The authors suggest that dextran reduces particle collisions, albumin forms a protective film, and sialic acid gives the vesicle surface a negative charge. Vesicles stored in DDAS can be administered intravenously directly after thawing. In an experiment on six young male mice with deep burns, stem cell vesicles stored in this mixture accelerated wound closure more than the same vesicles from a conventional buffer. By the 12th day, the wounds of mice that received vesicles from DDAS were almost closed. Freezing damages the membranes of vesicles and their cargo; DDAS, according to the authors, allows them to be administered without repeated purification.

🔗 Read original →

The enzyme CMLase removed age-related chemical damage from the amino acid lysine in proteins. On July 14, Nature Communications published a study on the CMLase enzyme, which removes CML, a common age-related chemical damage to lysine, from proteins and returns the original amino acid. The reaction was demonstrated on model proteins and human lens, skin, and aorta samples. Lens proteins and collagen in skin and vascular walls serve for years or decades. Reactive sugar and lipid compounds attach to their amino acids, resulting in glycation and lipoxidation products. One of these is Nε-carboxymethyllysine, or CML. The carboxymethyl group attaches to the side amino group of lysine and changes its charge from positive to negative. CML affects both the protein itself and the cells around it. It binds to RAGE, a cellular receptor for glycation products, and triggers inflammatory and oxidative signals. Cellular enzymes neutralize some of these reactive molecules even before they encounter the protein. CMLase works after CML formation, when the mark has already been fixed on long-lived proteins. The authors took a bacterial glycine oxidase and, over five cycles of directed evolution, tested more than 500 million variants. The resulting CMLase recognizes CML within an intact protein and converts the damaged residue back into lysine. On model proteins, after overnight treatment, the CML signal in an antibody test decreased by 52-97%. On soluble lens proteins from a 64-year-old donor, mass spectrometry showed a reduction in total CML of 45%, and the antibody test showed a reduction of 78%. Mass spectrometry measures the total amount of CML after chemical degradation of the protein, while the antibody test only sees the labels accessible to the antibody. The authors suggest that CMLase more easily removes labels on the surface of the protein than those hidden inside. On thin fixed sections of the aorta from a 75-year-old donor, CML staining decreased by more than 70%, and on skin from elderly donors, it decreased by more than 55%. The inactivated control enzyme retained the original signal. Soluble lens proteins and thin sections give the enzyme direct access to the target. In a living vessel, it needs to penetrate a dense extracellular matrix - a network of proteins that supports the tissue. Researchers also need to measure the immune response to the bacterial enzyme upon repeated administrations. CMLase allows for testing whether removing one chemical mark changes the RAGE signal and vascular elasticity. Glucospane, an age-related crosslink that connects matrix proteins, will require a separate enzyme. In the analysis of the glycation hypothesis, the accumulation of collagen and elastin damage is associated with the loss of elasticity in aging tissues.

🔗 Read original →

Brian Johnson was diagnosed with autoimmune gastritis, a disease in which the immune system attacks the stomach. On July 15, author on aging Andrew Steele broke down Johnson's public history: eleven years of low ferritin, an indicator of iron stores. A new team of doctors linked this signal to an autoimmune thyroid disease, checked the intestines and stomach, and biopsies confirmed autoimmune gastritis. In his account, Johnson describes how low ferritin was explained by a plant-based diet, training, sauna, and oxygen procedures. He tried food and supplements with iron, but the stores were not restored. Hemoglobin remained normal, so there was no anemia. "Low ferritin continued to be explained, but not corrected," Johnson wrote about the previous years. The new diagnosis began after a change of doctors. At 48, Johnson underwent a colonoscopy for the first time; the American preventive recommendation suggests such screening for people with average risk starting at 45. The colonoscopy ruled out hidden blood loss in the intestine. Then, doctors correlated the long-standing autoimmune thyroid disease with persistent iron deficiency, checked antibodies to stomach cells, and took five biopsies. The biopsies showed autoimmune gastritis: the immune system damages the stomach lining, which produces acid and helps absorb iron. In the early stages, this disease may not give a noticeable picture upon examination. The American Gastroenterological Association directly advises considering atrophic gastritis in unexplained iron deficiency and confirming it with histology - tissue examination under a microscope. In his analysis, Steele puts two facts from Johnson's text side by side: the Immortals Care protocol costing $1 million per year and the years when low ferritin was explained without looking for a cause. Sensors collect numbers; a diagnostic hypothesis links them to a cause and a test. On June 30, Steele compared personal protocols to a $45-70 million metformin trial. Such a project is looking for a general answer: does a cheap intervention work in people? Johnson's story adds medical consistency to this calculation: first, determine the cause of the disease, then test the treatment. Now, Johnson wants to move beyond observation. His plan starts with monitoring and support, then moves to influencing immune signals and regulatory T cells - cells that normally restrain the immune attack. In the long-term perspective, he considers CAR/CAAR-T: modified T cells that attempt to direct immune cells attacking specific tissue. He also writes about AI-designed antibodies and synthetic proteins. Johnson calls these steps research: part of the approaches still need to be created specifically for autoimmune gastritis. If one patient gets better after a complex set of interventions, the contribution of each of them, the natural course of the disease, and the role of usual supportive therapy remain unknown. A trial with a comparable group of patients separates the working method from a lucky coincidence.

🔗 Read original →

Weco reported that its agent rewrote the code of another agent and improved its results on external tests on July 14. Weco described AIDE²: over eight days, the system performed 100 cycles of rewriting and testing the internal AIDE agent. The company claims that during this time, seven consecutive improvements emerged, and two later versions outperformed the original on three external tests. The full technical report and the release of AIDE 85 itself, Weco promises later. The original AIDE is an agent for machine learning engineering tasks. It writes code variants, runs them, and develops successful branches. In AIDE², the external role was played by AIDE human - a manually tuned version of the same agent. It modified the code of the internal AIDE 0, a simplified version of the original AIDE. Each step included one edit and a full evaluation across task families. The version remained in the cycle only if a hidden check confirmed its advantage at a fixed evaluation cost budget. AIDE² optimized the solution search procedure itself. AIDE 85, on average, reduced the history of past attempts, which the model receives before the next step, by 16 times, and used the freed-up text volume for additional attempts. When the best branch stopped improving, the agent took its code as the basis for a new branch with a different strategy. This economy allowed it to conduct more experiments. A similar problem was already solved by AstraZeneca employees: their five agents passed a short map of past work to the next launch, and the reasons for the decisions were stored separately. In their tests, the system chose the correct model form in all 20 artificial tasks. Weco uses compressed history as part of a broader mechanism that checks on other task sets. Weco tested AIDE 47 and AIDE 85 on three external tests that the system had not seen in its improvement cycle. One of them is MLE-Bench Lite, a simplified version of MLE-Bench. The full MLE-Bench is compiled from 75 Kaggle machine learning engineering competitions and contains human benchmarks. In a separate test on KernelBench - a set of tasks for accelerating computations on a graphics processor - Weco considered a test bypass case when less than half of the acceleration claimed in the short test was preserved in the full workload. The share of such cases, according to the company, decreased from 63% in AIDE 0 to 34% in AIDE 85. Then the company placed the improved internal agent in an external cycle. On the tasks on which the system was improved, AIDE 47 reached the same maximum result in approximately 20 edits, and the manually tuned AIDE human - in approximately 40. This difference was within the statistical noise. The experiment showed the improvement of the internal agent, but did not confirm that it accelerates the next improvement cycle. Currently, Weco describes these results only in its corporate blog. The full technical report and AIDE 85 system, the company promises to release after completing the analysis. The authors also found that the statistical filter against test bypass in the later version was broken and actually had no effect on the result.

🔗 Read original →

GPT-5.6 helped to refute a 20-year-old hypothesis on checking thousands of scientific results at once. Statistician Edgar Dobriban found a data model where the Benjamini-Hochberg procedure slightly more often promises validity than it can provide. It is applied when one experiment simultaneously checks thousands of genes or other characteristics and it is necessary to limit the share of false findings. Dobriban published the proof and numerical certificate code; the model participated in the search, and the author checked the result. On July 14, Edgar Dobriban published a proof that concerns the usual protection against such coincidences. When a biologist compares thousands of genes between groups of cells, random coincidences are inevitable. Therefore, the results are usually passed through the Benjamini-Hochberg procedure: it orders p-values - numbers that show how much an observation resembles randomness - and selects a threshold so that the share of false alarms among the declared findings does not exceed the specified level on average. This procedure has long had clear guarantees for independent results and for some types of positive dependence. However, real genomic data is correlated: close DNA variants are inherited together, genes work in networks, and cells share common causes of changes. The question remained whether the usual procedure retains its guarantee for any correlated two-sided tests, where the effect can go in both directions. Dobriban built a specific model with a common hidden factor - one invisible cause that simultaneously shifts many results. In it, the distribution of null results and real signals changes. The procedure chooses a threshold at which there are slightly more errors among the selected findings than it promises. At a nominal level of 1%, a strict numerical certificate gives no less than 1.0416829%. The difference is small, but it breaks the universal hypothesis: for correlated two-sided data, the guarantee cannot be automatically transferred from the case of independent tests. On July 10, OpenAI published a candidate proof of a hypothesis on graphs; it still needs to be dissected by mathematicians. In Dobriban's case, the verification chain already includes an open numerical certificate. According to the statistician, GPT-5.6 received the mathematical formulation and in approximately 90 minutes generated a counterexample, the proof, and the certificate code. Dobriban checked the argument, and the open program recalculates the lower bound with interval arithmetic, where rounding goes to the safe side. Therefore, the result can be verified independently of the story about the model's capabilities. For the biology of aging, this is part of the scientific infrastructure. Mass measurements of genes, proteins, and cells constantly require distinguishing signal from random noise. AI that finds a verifiable counterexample helps to determine where the statistical method actually guarantees reliability and where its conditions have already ceased to be met. Source: Edgar Dobriban's manuscript, open code, and certificate.

🔗 Read original →

A wrist vibration and a program that learns together with a human have helped novices to more quickly control cursor signals from the brain. In a study of 31 people, the system simultaneously taught the user to represent movement and retrained the program that reads their EEG. Over several sessions, this combination gave a greater increase in accuracy than usual training or vibration alone. On July 15, in Nature Communications, a team from Carnegie Mellon University published a study on how to train a neurointerface: a system that reads the electrical activity of the brain and converts it into a command for a computer. In the experiment, a person looked at a cursor and imagined moving their left or right hand. An EEG helmet recorded the weak electrical signals on the skin of the head associated with the work of the sensorimotor cortex, and the program moved the cursor based on these signals. It is difficult for a novice to immediately create a stable signal that the program can distinguish from noise. The person and the program learn simultaneously: the algorithm changes the recognition rule while the user is looking for a clear way to control it. In April, a two-way neurointerface returned artificial sensation of walking to a participant through stimulation of the sensory cortex. There, feedback should accompany walking; here, the vibration on the wrist works as a training cue while the person learns to control the cursor. The authors gave them a common cue. A small motor on the wrist briefly vibrated during training attempts. After each block, the algorithm was updated and more strongly took into account attempts where the brain signal already well distinguished the desired direction. The vibration helped the person to find a reproducible signal, and the algorithm adjusted to this signal. The main group consisted of 15 people. Another eight received vibration, but the program did not highlight signals that the person found easier to master; eight worked in a usual system without vibration. Between the first attempt and the second session, the accuracy of control in the main group increased by 19%. In the group with vibration without such signal selection, the increase was 6.3%, and in the usual group, it was 3.0%. In a two-dimensional task, where the cursor moved on the screen in different directions, the average accuracy of continuous control reached 66.9%. A check with constant vibration and another check, where the program was updated without this connection, showed that the result arises from their joint work. In six participants of the main group, the advantage was preserved for more than two months, when the vibration was already removed. So far, this is a laboratory experience on healthy young people, not the restoration of movement in people after a stroke or spinal cord injury. In such patients, the sensorimotor cortex is often rearranged, and the paths of sensitivity can be disrupted. The speed of mastering the EEG neurointerface depends on whether the training of the person and the update of the algorithm coincide.

🔗 Read original →

In Shanghai, the first commercial implantation of NEO for restoring grip was performed. On July 13, doctors at Huashan Hospital implanted NEO in a patient who had been living with the consequences of a cervical spinal cord injury for ten years. During the operation, the system recorded stable signals from the brain's surface, and after the procedure, the patient's condition remained stable. Data on grip recovery, independence in daily life, and long-term safety for this patient have not been published yet. NEO is an implantable neurointerface that translates movement intention into a command for an external device. The NEO electrodes lie on the brain's hard shell, between the skull and brain tissue. The system recognizes the intention to clench a fist and sends a command to a pneumatic glove, which bends the fingers and helps hold an object. On March 13, the Chinese regulator registered NEO for restoring grip in adults with cervical spinal cord injuries. The registration allowed the sale and clinical use of the device. This operation became the first case where NEO was installed in a patient through a commercial route. According to SCMP, within four months after registration, the manufacturer started production, began working with hospitals, and selecting patients. The Shanghai supplementary medical insurance program Huahui Bao included implantation materials in its coverage: it reimburses 30% of eligible expenses, up to 150,000 yuan per year. The limit applies specifically to the materials, not the entire procedure.

🔗 Read original →

An implant in the sensory cortex has been restoring the sense of touch to people with spinal cord injuries for ten years. Five participants were implanted with microelectrode arrays - sets of thin electrodes - in the section of the sensory cortex associated with the hand. A study published on July 15 in Science Translational Medicine collected data over implantation periods of two to ten years: over 168 million electrical impulses and 27 cumulative years of device operation. To take a glass, a person must not only send a command to their hand to "squeeze fingers," but the brain must also constantly receive feedback from the hand: whether the fingers have touched a surface, how hard they are pressing, and whether the object is slipping. After a severe spinal cord injury, this feedback disappears. A person can control a robotic hand through a neurointerface and still not feel its contact with an object. Intracortical microstimulation delivers short electrical impulses directly to the sensory cortex - a section of the brain with a body map. Different points on this map are associated with different parts of the hand. When a microelectrode stimulates the corresponding point on the map, a person feels touch in the corresponding finger or palm. A sensor on a robotic hand can be connected to this electrode and convert contact with an object into artificial touch. Artificial touch has already been helpful in a task with a robotic hand. In a 2021 study, a participant with tetraplegia performed the task in an average of 20.9 seconds with only vision and in 10.2 seconds when artificial touch was added to vision. A neuroprosthesis for everyday use requires that the electrodes, brain tissue, and sensations remain functional for years. In a new study, the group of Charles Greenberg, Robert Gaunt, and Jennifer Collinger tracked this sensory channel for up to ten years. Each of the five participants was implanted with two electrode arrays in the sensory cortex. Over this time, the researchers delivered over 168 million impulses. They did not detect any serious complications associated with stimulation or signs that the impulses themselves were degrading the performance of the electrodes. The sensations continued to arise in the hand and remained localized. In one participant, after ten years, 60% of the electrodes were still working reliably; on average, across the group, 64% were working, with a spread of 13% between participants. A slowly increasing current was required for sensation: approximately 3.5 microamperes more per year. Rare sensations could briefly persist after the impulse was turned off, but did not require treatment. Two days ago, bioengineer Takeshi Kozaei called the preservation of living tissue around the electrode a condition for a neurointerface to last decades. In this study, part of the electrodes gradually lose sensitivity. A sensor on the hand must transmit a signal to the same point on the hand map in the cortex for years, and the system must maintain the sensation of contact while individual electrodes fail.

🔗 Read original →