science

Interesting read.

We show that large language models can be used to perform at-scale deanonymization. With full Internet access, our agent can re-identify Hacker News users and Anthropic Interviewer participants at high precision, given pseudonymous online profiles and conversations alone, matching what would take hours for a dedicated human investigator. We then design attacks for the closed-world setting. Given two databases of pseudonymous individuals, each containing unstructured text written by or about that individual, we implement a scalable attack pipeline that uses LLMs to: (1) extract identity-relevant features, (2) search for candidate matches via semantic embeddings, and (3) reason over top candidates to verify matches and reduce false positives. Compared to classical deanonymization work (e.g., on the Netflix prize) that required structured data, our approach works directly on raw user content across arbitrary platforms. We construct three datasets with known ground-truth data to evaluate our attacks. The first links Hacker News to LinkedIn profiles, using cross-platform references that appear in the profiles. Our second dataset matches users across Reddit movie discussion communities; and the third splits a single user's Reddit history in time to create two pseudonymous profiles to be matched. In each setting, LLM-based methods substantially outperform classical baselines, achieving up to 68% recall at 90% precision compared to near 0% for the best non-LLM method. Our results show that the practical obscurity protecting pseudonymous users online no longer holds and that threat models for online privacy need to be reconsidered.

A living drug, called CAR-T cell therapy, has revolutionized the treatment of blood cancers, achieving thousands of complete remissions of leukemias, lymphomas and myelomas since its first experimental use in 2010. The therapy involves extracting immune cells from the patient, genetically modifying them, and then reintroducing them, now with an enhanced capacity to destroy cancer cells. However, this successful treatment has so far failed against solid tumors, which are the most common kind. A new study released Thursday offers hope. An ultrasensitive version of CAR-T cell therapy has successfully eliminated human pancreatic, ovarian and kidney cancer tumors implanted in laboratory mice.

The lead researcher is the French-Canadian immunologist Michel Sadelain, born in Paris 66 years ago. His team demonstrated in 2003 that CAR-T cells, which target the CD19 protein on the surface of cancer cells, eliminated lymphomas in mice. Solid tumors—such as those of the breast, lung, colon, or pancreas—are more heterogeneous and do not display the CD19 protein, which has led to the failure of this strategy over the past two decades.

Sadelain’s group at Columbia University in New York is proposing a new target: the CD70 protein, characteristic of more than 20 types of solid tumors. It’s a well-known target, but until now no experimental treatment had worked because it’s apparently present in only a percentage of the cancer cells within a given tumor. One of the researchers on the Columbia team, Sophie Hanina, had a hunch: perhaps the CD70 protein was indeed present in all tumor cells, but in some of them at such minute levels that it was undetectable by standard CAR-T cell therapies.

Schrödinger first proposed in the 1920s that color perception could be mapped in a three-dimensional space. He suggested that the eye’s three types of light-sensitive cells — often linked to red, green, and blue — shape how people experience hue, saturation, and lightness. His concept helped scientists think of color not just as a physical property of light, but as something organized inside the visual system.

For decades, the theory influenced research in vision science and color studies. Yet key parts of the framework were never fully defined. One major gap involved the so-called neutral axis — the range of gray shades stretching from black to white. Schrödinger referred to this axis but did not clearly define it, leaving an important question unanswered. Researchers say that this gap has now been closed.

Reversing Paralysis? Not nearly on the market yet, but ' Lab Grown Organoids' are highly-promising

Link to paper: https://www.nature.com/articles/s41551-025-01606-2

If the basic logic of the DSM is flawed, it should be abandoned. Instead, psychiatrists should move towards a system that looks at an individual’s mental experiences in context, alongside their unique developmental vulnerabilities and strengths, as the main source for analysing and responding to their distress. Diagnosis would no longer name a disorder but map what kinds of support, relationships and learning processes are most likely to help a person regain agency, coherence and a sense of future.

• The number of people reporting generalized anxiety disorder and major depressive episode nearly doubled, increasing to 5.2 per cent and 7.6 per cent, respectively.
• Suicidality didn’t change much in adults but increased by 44 per cent among youth. Younger Canadians also saw some of the strongest cannabis–mental health connections.
• The number of people using cannabis multiple times a week more than doubled, while the number of people who reported using cannabis in the last year increased to 20.7 per cent.
• Canadians who used cannabis at any level, compared to those who did not, were more likely to meet criteria for generalized anxiety disorder and major depressive episode and report suicidality. The connection between cannabis use and these mental health problems strengthened over time.
• In 2022, Canadians who used cannabis regularly (two or more times a week) were about five times more likely to report anxiety, depression, or suicidality than those who did not use cannabis.

Research.

People who prefer structured, rule‑based explanations may find conspiracy theories appealing because they offer a clear, ordered explanation for events that feel chaotic. New research has found that understanding how someone processes information can be a strong predictor of whether they are drawn to conspiracy beliefs.