Science
Original Science analysis and news, written and editorially gated by the Omega team.

Pat Gelsinger Bets on Light-Based Lithography to Revive Moore's Law
The former Intel CEO explains why advanced chip-printing technology—not processor design—holds the key to semiconductor breakthroughs.

Heavy AI Reliance Lowers Test Scores Despite Higher Satisfaction
New research reveals a paradox: students who depend most on generative AI feel they learn more but perform worse on assessments.

AI Detects Pancreatic Cancer 16 Months Before Doctors Can
Johns Hopkins surgeon says machine learning models identify deadly cancer patterns that take physicians decades to recognize, though clinical validation remains early.

AI Index 2026: Safety Research Lags as Models Grow More Powerful
Stanford's annual benchmark report reveals productivity gains across industries but warns oversight isn't keeping pace with capability advances.
Inductive Bio wins drug metabolism AI contest in 28-way tie
OpenADMET competition reveals that training data quality may matter more than model architecture for predicting how the body breaks down drugs.
USC Researchers Use Sweat, Brain Signals, and Eye Tracking to Detect Depression
A DARPA-funded study combines multiple biomarkers with AI to create objective mental health assessments that could supplement traditional clinical interviews.

Inference Bottlenecks Push Optical Interconnects Into AI Chips
As AI workloads shift from training to deployment, researchers are integrating photonics directly into processor packages to overcome bandwidth and power constraints.

Mayo Clinic Sued Over AI Research Safety and Oversight Practices
Former research operations director claims she was fired for raising concerns about institutional review board manipulation and patient data privacy.

AI Accelerates Superconductor Discovery, Finds Two New Materials
Machine learning pre-screening helped researchers identify and synthesize YRu3B2 and LuRu3B2, demonstrating a faster path to room-temperature superconductors.
AI Agents Use 136x More Energy Than Chatbots—But Idle Time Is the Real Cost
A KAIST study reveals that agentic AI's power problem isn't just consumption—it's expensive GPUs sitting idle up to 54% of the time while waiting for external responses.

Gen AI creativity plateaus without 'idea co-development'
New research shows workers must learn to refine AI suggestions, not just request more of them, to sustain creative gains over time.

Medical ethics journal retracts high schooler's AI paper
Fabricated references and peer review manipulation led to the withdrawal of a pharmaceutical AI ethics study.

AI Tools Are Eroding Professional Skills, New Studies Show
Research in medicine and software engineering reveals measurable declines in human expertise when workers rely on artificial intelligence systems.
AI Drug Discovery Excels at Molecules, Fails at Clinical Trials
Billions invested in AI pharma have accelerated early-stage chemistry but left the costliest phase—proving drugs work in humans—largely untouched.
Aquifer Thermal Storage Could Cut Data Center Cooling Costs
University of Illinois researchers propose using underground water systems as natural thermal batteries to reduce energy and water consumption in AI facilities.
China's LineShine supercomputer tops global rankings
The system achieves 2,000 exaflops using CPU-only architecture, circumventing US chip export restrictions.

Neuroscience Foundation Models Need Standardized Data First
Recent AI breakthroughs in brain research succeeded not because algorithms improved, but because researchers spent years making data compatible.

Lancet Correspondence Challenges AI Mammography Evidence Standard
A formal response to the MASAI trial argues that sensitivity gains alone don't justify practice recommendations when overdiagnosis signals remain unaddressed.
Physicist challenges Microsoft's topological qubit claims
A peer-reviewed Nature paper argues the Majorana 1 chip did not demonstrate a working topological qubit as the company claimed in 2025.
AI Search Systems Now Training on Their Own Output
A feedback loop is emerging as AI models increasingly consume AI-generated content to formulate answers, potentially narrowing information diversity.

USC Researchers Stress-Test AI Chatbots for Mental Health Care
A new study enlisted 100 mental health professionals to evaluate how leading language models respond to real patient questions, revealing both promise and safety concerns.
ORNL Advances Autonomous Labs with AI-Driven Decision-Making
Rob Moore explains how artificial intelligence is evolving from automation to true autonomy in scientific research at Oak Ridge National Laboratory.
Nexentis subsidiary partners with Boltz AI to screen drug targets
MitoCareX Bio will combine proprietary structural modeling with foundation models to identify small molecules against transporter proteins.

Space-Based Solar Power Emerges as Answer to AI Energy Crunch
With data center electricity demand set to more than double by 2030, companies are looking beyond Earth's grid to continuous orbital energy sources.