Payments experts say a potential Stripe–PayPal deal could accelerate digital payments innovation, but regulatory hurdles, integration complexity and valuation remain the biggest obstacles.
Get the latest international news and world events from around the world.
Anthropic Reportedly in Talks to Acquire Peer Decart for $6 Billion, Bolstering Compute Ahead of IPO
AI startup Anthropic is reportedly in talks to acquire peer Decart AI for approximately $6 billion (about NT$190 billion). If finalized, the deal would mark the largest acquisition in Anthropic’s history, coming at a time when the company is preparing for its highly anticipated IPO. Bloomberg first reported the deal on August 13, noting that the agreement has not yet been finalized and negotiations could still fall through.
Anthropic has been moving at a rapid pace on both fundraising and listing preparations this year. In February, it completed a $30 billion Series G round at a post-money valuation of $380 billion. Less than three months later, in May, it closed a $65 billion Series H round, pushing its post-money valuation to $965 billion — nearly $1 trillion. In early June, Anthropic confidentially filed its S-1 registration statement with the U.S. Securities and Exchange Commission (SEC), paving the way for a future public listing. If this acquisition goes through, it would occur during a period when the company’s valuation and IPO preparations are both accelerating simultaneously.
Market sources indicate that Anthropic, which has historically made very few large acquisitions, has been investing heavily in compute capacity to develop new products and serve customers. According to sources, Decart’s software enables chips to operate more efficiently, thereby reducing the cost of training AI models. This technology would help Anthropic’s existing infrastructure handle greater demand. Upon completion of the deal, Decart’s team would join Anthropic’s inference and performance division. According to Forbes, FierceBiotech, and other media tallies, Anthropic has completed at least four acquisitions this year, including compiler startup Stainless and biotech startup Coefficient Bio (in a roughly $400 million stock deal), but all were far smaller than the Decart transaction — $6 billion would be more than ten times the size of any previously known deal.
Thermal detection of single photons using Dirac fermions
Single photon detectors are essential for various quantum and imaging applications. Here, the authors report graphene bolometers able to detect single near-infrared photons at temperatures up to 1.2 K with intrinsic quantum efficiency up to 87%, dark count < 1 per second and effective noise equivalent power down to 2 × 10−22 W/ $$\sqrt{{{{\rm{Hz}}}}}$$ Hz.
Distinguishing true progression from treatment effects in glioblastoma: a practical, evidencegraded imaging framework for the multidisciplinary team NeuroOncology
Distinguishing true progression (TP) from treatment effects—pseudoprogression (PsP) and radiation necrosis (RN)—after chemoradiation for glioblastoma (GBM) is a consequential, unresolved decision that conventional MRI cannot reliably make in 30–40% of cases, and that is rarely made by any single specialty alone.
We synthesized and graded evidence from RANO 2.0, advanced MRI (perfusion, diffusion, and spectroscopy), amino acid PET (per PET RANO 1.0), and radiomics, based on a literature search of PubMed/MEDLINE, Embase, and Cochrane (2010–2025).
In selected cohorts, combined MRI plus amino acid PET reports areas under the curve of 0.90–0.95 versus 0.65–0.72 for conventional MRI alone; however, these figures come from cohorts spanning the full range of post-treatment enhancement—including easily classified cases—rather than the ambiguous subset in which advanced imaging is actually used, and therefore likely overstate real-world accuracy. We organize the evidence around clinical modifiers of pretest probability—MGMT methylation, interval since chemoradiation, neurologic trajectory, antiangiogenic or immunotherapy exposure, reirradiation, and lesion location—that determine how heavily each imaging tier should be weighted. For amino acid PET, we address U.S. access, including recently published prospective and multicenter diagnostic-accuracy data for 18 F-fluciclovine.
New brain map connects mice and primates to strengthen medical research
Most neuroscience research still runs on mice, even though much of what applies to mice does not hold up in humans. Marmosets have emerged as a bridge between species. However, comparing the brains of mice and marmosets runs into a chicken-and-egg problem. Scientists need to know which part of one brain matches the respective region of the other, but almost every brain map draws those areas differently.
For a new study published in Communications Biology, Cold Spring Harbor Laboratory Professor Partha Mitra and colleagues built the first cross-species map to systematically link mouse and marmoset brain regions.
“It’s the kind of problem that nobody wants to address, because it’s hard,” Mitra says. Comparing genomes across species is relatively straightforward. DNA aligns letter by letter. But brain circuits vary even within one species. Compounding the problem, research teams have built different brain atlases that divide the same mouse brain differently.
PeptideMHCtargeted engineered viruslike particles enable selective priming and gene editing of tumorspecific T cells
As well as T cells bearing an engineered TCR55-A50E, which forms catch-bonds with both HIVpol/B∗35 pMHC and Pep20/B∗35 pMHC. We treated TCR55 and TCR55-A50E cells with pMHC-eVLPs pseudotyped with either HIV/B∗35 or Pep20/B∗35 and subsequently assessed T cell activation by CD69 upregulation and eVLP transduction. Consistent with previous reports, we observed increased CD69 upregulation in TCR55 cells by Pep20/B∗35 eVLPs compared to HIV/B∗35 eVLPs, whereas TCR55-A50E cells were similarly activated by both pMHC-eVLPs (Figure S4G). Notably, pMHC-eVLPs presenting HIV/B∗35 were still able to induce CD69 upregulation on TCR55 cells despite reported slip-bond formation (Figure S4G). Interestingly, both HIV/B∗35 and Pep20/B∗35 pMHC-eVLPs were able to efficiently transduce both TCR55 and TCR55-A50E cells (Figure S4H), suggesting that antigen specificity and physiological TCR·pMHC affinity, rather than catch-bond formation, are the primary determinants of pMHC-eVLP-mediated T cell activation and transduction.
Collectively, these findings demonstrate that the pMHC-eVLP platform can be programmed with different antigens and HLA allotypes to enable selective and efficient targeting of tumor-specific T cells via natural TCRs with physiological antigen affinities, and that pMHC-eVLPs induce antigen-specific T cell activation during TCR-mediated entry.
Sleep could help identify people at risk of developing Alzheimer’s disease at an earlier stage
What if certain signs of the disease were to manifest subtly during sleep, long before the first memory problems appear? This is the line of inquiry being explored by a team of researchers at ULiège.
A team of scientists from the University of Liège (GIGA Neurosciences), supported by the Stop Alzheimer’s Foundation, has analyzed the sleep patterns of more than 500 healthy people. Among middle-aged participants, a higher frequency of nocturnal micro-awakenings was found to be associated with a greater genetic risk of developing Alzheimer’s disease, whereas this link was not observed in young adults. This research, published in the journal Sleep, suggests that the study of sleep could, in the long term, contribute to the early identification of vulnerable individuals.
