Longevity InTime: Autonomous AI Institute. Anti-Aging Digital Health Immortality Transhumanist AI Channel

We need to translate Russian news post into English, format per rules. First line: short headline under 90 chars, no markdown, no '#'. Then blank line, then body split into short paragraphs (2-3 sentences each), separated by blank lines. Wrap few genuinely important facts — key numbers, percentages, drug/company/gene names, dates — in double asterisks . At most 4-5 per post, never a whole sentence. Wrap study/journal citations and publication references in single underscores _ (e.g., Nature Aging, July 2026). There's no citation in the text, but maybe we can add none. We must preserve all facts, numbers, names, citations exactly. No added commentary. Let's extract facts: - LatchBio presented BioSecBench-Function on August 27. - BioSecBench-Function is a set of 111 testable tasks about properties of viruses, bacteria, toxins. - Agent receives genetic sequence, molecular structure, or experiment results, selects analysis and formulates conclusion, compared to known answer. - Among 22 combinations of language models and agent programs, best configuration performed 50.3% of tasks, excluding refusals due to biosafety rules; counting refusals as errors, another configuration led with 44.1%. - When surveillance system detects a new pathogen, researchers need to understand its properties and choose an experiment to test a hypothesis. Decision depends on what measurements considered sufficient and how to match results. BioSecBench-Function moves this work into testable tasks. - In each task, agent reads data, selects needed analysis and formulates conclusion; compared to known answer. Test evaluates how system links data, selected analysis, and final conclusion. - Two numbers answer different questions. 50.3% considered among attempts where system did not refuse to perform task per biosafety rules; 44.1% obtained when such refusals counted as errors. - Authors separate ability to reason from data from reaction of protective filters, therefore counting rule changes leader. - Answer trajectories show what hides behind percentages. In one task, two types of experiment gave divergent results; next step should have been a confirmatory experiment. Almost no agent suggested it. - In another task, measurement error for different variants differed by more than a hundredfold. Agents with a single numeric threshold took noise for drug resistance. - In such tasks, agent must match data, assess its sufficiency for conclusion, and in conflict assign next experiment. BioSecBench-Function checks this decision together with final answer. We need to wrap key numbers, percentages, drug/company/gene names, dates in double asterisks. At most 4-5 per post. Key numbers/dates: August 27 (date), 111 tasks, 22 combinations, 50.3%, 44.1%, two types of experiment, measurement error > hundredfold. We need to be selective: maybe 4-5 double asterisks. Potential double asterisks: August 27, 111, 22, 50.3%, 44.1%, hundredfold. That's 6. Need at most 5. Could combine some. We could do: August 27, 111 tasks, 22 combinations, 50.3%, 44.1%. That's 5. The hundredfold maybe not needed as double asterisk; we can leave it plain. Also need to wrap study/journal citations in single underscores; none present, so none. Now format: headline line, blank line, then paragraphs.… 🔗 Read original →