Practical Magic

A 3D-printed concrete shell can stay standing forever without a single piece of steel reinforcement thanks to an AI that solved a 350-year-old math problem.

An AI can build a perfect 3D model of a child's spine using only MRI data, completely removing the need for dangerous X-rays.

Artificial intelligence can watch a video of cancer cells and write down the fundamental mathematical laws governing their growth.

A common heart failure medication can lock HIV inside immune cells and prevent the virus from waking up.

Electric cars can now predict a driver's next move with 90% accuracy a full second before the person actually turns the wheel.

A new way to sort lithium isotopes using liquid electrolytes could solve the fuel crisis for future nuclear fusion reactors.

Listening to a lithium-ion battery with ultrasound can give you a 15-minute warning before it explodes.

A chip-scale nuclear clock could be a hundred times more precise than the atomic clocks used to run GPS today.

The structural scaffolding inside our cells can be programmed like photonic hardware to control how they interact with light.

A common amino acid used by the human body can be turned into a crystal that outperforms the high-tech materials in your phone.

A brain oxygen level below 46% during CPR marks a physiological point of no return where resuscitation is no longer possible.

A graphene based membrane uses simple pH changes as a light switch to pull lithium out of old batteries with perfect accuracy.

A nanoscopic metaatom can be switched on and off with electricity to create holograms that can be rewritten instantly.

A 6.2-Gigawatt tabletop laser packs the power of several nuclear reactors into a device small enough to fit in a standard lab.

High frequency magnetic fields can weave liquid hydrogels into complex, organized tissue structures inside a living body.

A new chemical reactor cheats the laws of thermodynamics to produce carbon monoxide at temperatures once thought impossible.

A crossed bowtie antenna can expand a quantum sensor's detection area by 15 times without adding any noise or defects.

Current computer simulations of organic thin films are getting the arrangement of molecules wrong even when they get the average correct.

A quantum framework for network analysis can count complex patterns in massive datasets using exponentially fewer resources than a supercomputer.

An AI agent named LM-Tree can out-calculate human editors by 40% when deciding exactly how much a news article is worth to a search engine.

A new class of control pulses can manipulate nuclear spins 30 times faster and more accurately than any previous method.

Wild animals can be fitted with GPS tracking collars without ever being touched or tranquilized by a human.

A new recycling process can turn common biodegradable plastic waste into a more valuable, high-performance material.

A new biosensor made of high entropy alloys is more sensitive than a PCR test at detecting ovarian cancer.

A new chemical reaction uses alternating current to tag pharmaceutical molecules with heavy hydrogen for easier drug testing.

A new mathematical tool can finally sort the chaotic vibrations of glass into organized maps.

An automated pipeline of AI agents discovered 118 real-world security holes and 203 zero-day vulnerabilities in a single run.

Six brand new mathematical discoveries were generated by an autonomous multi-agent system with zero human intervention.

Waste graphite from old lithium-ion batteries was hit with a millisecond pulse of heat and became 12% more efficient than brand-new material.

A multi-agent AI pipeline successfully found real-world security flaws in the ISO C++ standard that human experts missed for years.

Spiking neural networks replace energy-hungry matrix multiplications with simple additions to run large language models.

Capability-sealed tokens allow AI agents to use API keys without ever knowing the actual secret.

A new hardware component called a neuristor uses a metal-to-insulator transition to mimic the way the human brain shuts down signals.

Classical computers can now simulate quantum circuits that were previously thought to be impossible without a quantum machine.

Reverse-engineered executable specifications allow AI to fix 94% of software bugs that would normally stump a human.

A 7B parameter model solved more formal math theorems than a 671B parameter giant by using a small Guide model to police its own reasoning.

A new brain-mimicking computer chip can train on complex data in milliseconds, a task that takes standard GPUs several hours.

AI models can now think harder and improve their own answers on the fly by spending more compute time on a specific question.

A formal verification engine named COBALT uses Z3 logic to find arithmetic bugs in the C++ walls that keep AI trapped in its sandbox.

A small 2 billion parameter model generates better video than a rival seven times its size.

A digital lock now makes a top-tier AI model fully useful to its owners but impossible for anyone else to steal via distillation.

A specific class of network dynamics is computationally impossible for classical computers but takes only seconds for a quantum algorithm.

Magnetic fluctuations in a yttrium iron garnet crystal generate perfect randomness at speeds of 20 megabits per second.

A small 3 billion parameter model can learn to say 'I don't know' and explain what information is missing.

The standard best practice for training the world's most powerful AI models actually hurts their performance in certain situations.

An automated system generates C++ code that runs nearly 10 times faster than the best work of human performance engineers.

Thirty minutes is all it takes to teach a humanoid robot to pick up a brand-new object using foundation models.

A new scheduling framework allows a single consumer GPU to process billion-token sequences that would normally require a supercomputer.

A new translation method gives you the high quality of slow-thinking AI models at the lightning speed of fast-thinking ones.

A new class of mathematical functions represents deep neural networks with polynomial complexity instead of exponential size.