A curated list for Self-Improvement in Foundation Model Based Agentic Systems. — selfimproving-agent/Awesome-Self-Improving-Agents
For decades, pathologists have diagnosed and graded breast cancer by looking at tissue samples under a microscope, searching for telltale signs of disorder in the structure of cells and tissues. Now, researchers at Columbia and their collaborators have developed a new computational approach that transforms those visual patterns into quantitative measurements, potentially improving how clinicians predict breast cancer outcomes and choose therapies.
In a recent study published in Cancer Research, researchers used mathematical tools known as topology to develop biomarkers quantifying the organizational structure of breast cancer tissue. The approach generated continuous numerical scores that predicted patient survival and treatment response more accurately than many traditional biomarkers, while also showing less variation across racial and ethnic groups.
Higher baseline plasma p-tau217 was associated with greater 5- and 10-year risks of cognitive impairment and faster cognitive decline among 2,684 cognitively unimpaired adults. Very high levels corresponded to an estimated 38% risk at five years, but selected cohorts and limited long-term data mean routine screening and individualized prediction remain premature.
David Szymanski, creator of Dusk and one of the chief architects of the retro FPS revival, grew up playing games like Lemmings and Oil’s Well on the family PC. Among these family favourites was a formative first-person game. But it probably isn’t the one you’d expect: “The most foundational one I can remember is playing Myst for the first time,” says the New Blood designer. “That was the first game I remember really getting super into—being able to enter another world and explore it in first-person.”
Initially a designer of short horror games, Szymanski’s breakout came in 2018 with Dusk, an homage to classic ’90s shooters like Doom and Quake. Following Dusk’s success, Szymanski returned to developing shorter games, including the hit submarine horror game Iron Lung, which was adapted into a movie by YouTuber Markiplier earlier this year.
Efficient positron injection in plasma wakefield acceleration is challenging because it requires a field structure that can both focus and accelerate positrons. Here, we propose a self-adaptive catch-up injection scheme, in which a positron source placed in front of a hollow, positively charged driver is naturally captured as the wake catches up. Analytical modeling and 3D PIC simulations show that positrons with broad initial conditions can be injected into similar accelerating phases and further accelerated to multi-GeV energies. These results provide a proof-of-principle demonstration for a new positron injection and acceleration scheme in plasma wakefields.
From the beloved vanilla orchid, whose pods flavor many of our favorite desserts, to the nose-twisting corpse flower, whose gigantic flower draws crowds with every fetid bloom, even the world’s most treasured flowering plants are in grave danger.
And scientists are only just coming to terms with just how much we stand to lose.
In a recent paper published in Science, an international team of biologists warns that we risk losing more than a fifth of the evolutionary history of flowering plants, as a result of ever-increasing human pressures.
Blood transfusions are crucial in human and veterinary medicine, yet human blood reserves are in constant need of donors, while blood bank systems in veterinary care are nearly nonexistent. This leaves canine transfusions largely reliant on donations from healthy dogs, but securing compatible blood remains a major challenge because dogs also have different blood types.
Fortunately, recent advances in induced pluripotent stem cells (iPSCs) have created new opportunities for producing blood cells in the laboratory. This and the similarities between human and canine health have drawn attention to dogs as translational models that can bring about innovative changes in both fields. However, a method for generating red blood cells from canine iPSCs has not been fully established.
On Thursday, the singer Ella “Lorde” O’Connor took to Instagram to share a screenshot of the “About the Song” summary of one of her songs. In the screenshot, the AI feature can be seen declaring that “on her Ultrasound World Tour, Lorde turns Current Affairs,” a song off of O’Connor’s most recent album, “into a full-on performance piece, stripping down to underwear while a dancer pours water over her stomach so the song plays out like the shower scene she talks about on stage.” According to a blurb at the bottom of the summary, it drew information from the Australian outlet The Music, which reviewed O’Connor’s tour.
But according to O’Connor’s Instagram story, this isn’t true — that performance happens in her show, but in a different song. There was an inaccuracy in The Music’s review, and that made its way into Spotify’s AI summary.
“Not only is this inaccurate (not the song I did that in),” O’Connor wrote in the caption of the Instagram story, “but reducing a song to an AI-generated meaning right at the source feels like it limits free interpretation [in my opinion]. At least make it possible for artists to opt out please.”