RealFeedApp beta
Top science & technology news, August 15, 2026
- Semaglutide linked to 26% lower 5-year predicted dementia risk
Study finds semaglutide may reduce the 5-year predicted risk of dementia by 26%.
- Auto-research with codex: How I achieved a 232x Faster Kernel
Author describes how they used the Codex model to speed up kernel analysis and achieve significant results.
- Second Near-Infrared Window Squeezed Light-Field Microscopy Enables High-Speed Volumetric Imaging of Deep-Tissue Dynamics in Vivo
High-speed three-dimensional imaging in scattering tissues remains challenging because volumetric microscopy generally requires scanning, whereas snapshot light-field approaches divide limited detector pixels among multiple views. This constraint is particularly severe in the second near-infrared window (NIR-II), where commonly used InGaAs cameras typically have relatively small sensor formats and high detector noise.
- SpaceX officially closes its Cursor acquisition
SpaceX has finalized the acquisition of AI startup Cursor, as announced on their blog. The deal was made after both companies began working together to develop technologies.
- Every fusion startup that has raised over $100M
Despite initial skepticism, fusion power is gaining traction and attracting significant investment. If successful, these startups could revolutionize energy markets.
- A CRISPR-Cas9 platform for primary human hepatocytes enables arrayed screening and in vivo validation of HBV host factors
Researchers have developed a CRISPR-Cas9 system that can efficiently edit genes in primary human liver cells, allowing for both in vitro screening and in vivo validation of hepatitis B virus (HBV) host factors. This breakthrough could lead to new treatments for liver-related conditions.
- v3.1.0 of the OpenAI Python SDK
The OpenAI Python SDK released version v3.1.0 with new features and improvements.
- Ukraine Strikes Major Russian Rocket Factory with Cruise Missiles
Ukrainian President Volodymyr Zelenskyy confirmed that his country struck the Progress Rocket Space Center deep within Russia using ground-launched cruise missiles. The largest city in the Samara Oblast, about 900 km from the Ukrainian border, is home to Russia’s most important manufacturer of rockets.
- Practice Makes Unsafe: Skill Misevolution in Self-Improving LLM Agents
The paper explores how skill evolution can make unsafe successes reusable, potentially leading to mishandled experiences and mismeasured skills.
- Show HN: ThoughtDAG – An editable context graph for LLM conversations
ThoughtDAG is a tool that allows editing and analyzing the history of interactions between users and Large Language Models (LLMs). It helps prevent repetition by allowing new requests to be created without overwriting previous conversation histories.
- REM Sleep Disengagement of β Oscillations Permits Rapid Dream Movements in Parkinsonism
REM sleep behavior disorder is a hallmark of prodromal -synucleinopathies, yet why patients with Parkinson's disease can generate rapid, coordinated movements during REM sleep despite daytime bradykinesia remains unknown. Here we combined recordings of eye movements, cortical electroencephalography, and basal ganglia-thalamic neuronal activity across vigilance states in non-human primates before and after MPTP-induced parkinsonism. Parkinsonism enhanced {beta} oscillations and impaired movement-related neural dynamics during wakefulness and NREM sleep, whereas both pathological {beta} activity
- Can LLMs work in the wet lab?
HackerNews AI publishes results from BenchBench-Protocol, a benchmark built from thousands of real-world experiments that tests whether models can troubleshoot and optimize wet lab protocols like expert scientists.
- Skye and Redx Form New Entity to Advance Fibrosis Assets with $125M Financing
Fibrx Therapeutics will receive $125 million in financing from the new public company formed by Skye Bioscience and privately held fibrotic disease biotech Redx Pharma, combining their assets for fibrosis research.
- Diagnostic Foundation for Evaluating LLMs' Research Integrity as Co-Scientists
The authors introduce IntegrityBench, a benchmark to evaluate misconduct classification, ethical reasoning, and decision-making grounded in artifacts across 36 paired tasks under a five-level implicit-explicit pressure protocol spanning three domains and four research stages.
- An Anecdote Against Slop Artifacts
The author shared a cautionary tale about their recent paper on verifying exact samplers for continuous distributions.
- Single-Cell Profiling Reveals Tumor Grade-Dependent Immune Remodeling in BRAFV600E-Driven Glioneuronal Tumors
BRAFV600E is the key driver variant in epilepsy-associated glioneuronal tumors (GNTs). These tumors often share MAPK/PI3K hyperactivation, a generally benign biological course and rare occurrence of malignant variants.
- Focused framework sampling recovers binding-positive humanized anti-amyloid-beta antibodies in a single sorting round
Therapeutic antibody development requires workflows that integrate antigen-reactive clone discovery with efficient humanization and early developability assessment. Here, we combined immune yeast fragment antigen-binding (Fab) display with single-round focused humanization and applied the workflow to antibodies against amyloid-beta (Abeta)-derived preparations. Immunization with Abeta1-42 aggregate preparations generated a Fab-display library with a diversity of approximately 3.5 x 10^8. Magnetic enrichment followed by fluorescence-activated cell sorting (FACS) identified three sequence-distinct clones.
- CDMOs Expand and Build for Future Growth
Executives from Fujifilm Biotechnologies, Lonza Group, MilliporeSigma, Samsung Biologics, and Thermo Fisher Scientific discuss their companies' recent service expansions and new facilities in interviews with GEN.
- Improving system safety with Temporal Logic of Actions (TLA+)
The hardest distributed systems bugs to find aren't in any single operation. Two processes each do the right thing, but in an order nobody thought of, and data disappears. Tests don't catch these bugs because tests run the interleavings you imagined, while the bug is in the one you didn't.
- v0.32.13
An error occurred while loading version v0.32.13, please reload the page.
