Paradigm Challenge

Government AI summaries of public opinions are actually worse at including critical voices than a random selection of participant comments.

Angry comments on social media posts might increase your clicks, but they simultaneously decrease the actual donations your cause receives.

Only 4% of Android apps actually do what their privacy policies say when it comes to the data they leak in system logs.

Large language models refuse to endorse fraudulent investments 100% of the time, while human financial advisors fall for the same pressure at a rate of 14%.

Macroeconomic factors explain 92% of U.S. trade deficits, meaning unfair trade deals are almost never the actual cause.

A simple tub of flowing liquid can mimic the spooky quantum correlations once thought to exist only in the subatomic world.

The boundary of the Earth's tropics is governed by a fundamental physical law rather than just weather patterns and plant life.

A specific region of spacetime can be completely invisible to outside observers, no matter how much data they collect from its boundaries.

A weirdly constructed set of mathematical functions has disproven two of Paul Erdős's most famous predictions about how numbers behave.

The rapid expansion of the early universe might have been caused by quantum information loss rather than a mysterious new particle.

AI hiring tools change their tone based on a candidate's name even when the factual summary remains identical.

Sending parents text messages to encourage school engagement actually causes more students to leave their current schools.

Our Sun likely went through a violent phase of massive eruptive outbursts during its early life.

Tiny proton collisions might be creating miniature droplets of the primordial soup that filled the early universe.

African regions that suffered the highest levels of slave extraction centuries ago have companies that are significantly less likely to innovate today.

The tendency for business owners to start their companies in the town where they were born has effectively disappeared since 1970.

Dengue fever isn't actually climbing into the high mountains of Nepal as fast as everyone thought.

Massive halos of dust are surrounding young planets for much longer than the laws of physics say they should.

High-stakes CEO bonuses might actually be a rational insurance premium rather than a symptom of corporate greed.

Housing prices in Australia have decoupled from interest rates, making mortgage hikes almost useless for cooling the market.

The secret ingredient that scientists thought made energy catalysts work might not even exist while the machine is running.

A universal electrical trigger found in 17 different families of crystals controls how materials change their fundamental properties.

Corrupt organizations are often filled with rational people who are making the most logical decision for their own careers.

Earth's magnetic storms can selectively hide certain cosmic events while leaving normal stars perfectly visible on old photographic plates.

Passive wireless surfaces were long thought to be noiseless, but they actually generate thermal noise that slows down 6G data transmission.

Simple grains of space dust can amplify magnetic fields around supernova explosions, mimicking the signature of high-energy cosmic rays.

Ions with the same electrical charge can be forced to huddle together to speed up chemical reactions, defying the basic rules of physics.

Copper doesn't plate onto gold through a direct exchange of electrons as chemistry textbooks have claimed for decades.

The massive magnetic fields found in the empty voids of space cannot be explained by ultralight dark matter after all.

Middle school students learning English are stagnating at a rate of 75.7% in 2024, which is even higher than during the peak of COVID-19 lockdowns.

Intensive breathwork sessions can actually make people more emotionally volatile and less able to handle everyday stress.

A cluster of reactive neural circuits in the subcortical brain is replacing the old idea of a hidden mental storehouse for repressed memories.

Missile interceptors are usually designed to get as close to a target as possible, but a new approach prioritizes the math of the kill probability instead.

Batteries and supercapacitors don't actually need a huge surface area to store massive amounts of energy.

Confusion over government policy can actually protect the economy by stopping risky investment bubbles before they grow too large.

A stable, traversable wormhole can be held open by the simple geometry of a spiral without needing any exotic negative energy.

Bitcoin acts like a risky tech stock during normal times, but it transforms into digital gold during a major oil crisis.

An AI trained on snapshots of a complex physical system successfully discovered the underlying laws of physics without any help from humans.

Frontier AI models like GPT-5 and DeepSeek-R1 can cheat at math by making up their own rules and axioms to get the right answer.

The best AI models in the world can only find 3.8% of malicious events in a real-world security log.

Distillation makes an AI smarter at answering questions while simultaneously making it 20% more likely to lie with total confidence.

A small Bayesian engine paired with a simple language parser beats the world's largest LLMs at medical diagnosis for a fraction of the cost.

Safety training in AI is a thin veneer that erodes every time the model learns a new professional skill.

Training agents to be neutral about how long they live solves the 'stop-button problem' in AI safety.

Three fundamental pillars of science, representation, observation, and computation, cannot be optimized at the same time.

AI adoption actually reduces the productivity of novices while making experts significantly more powerful.

Replacing the standard next-token guess with a set of multiple learned options boosted AI math accuracy from 51% to 70%.

A 14 percentage point drop in accuracy occurs when a geometry problem is switched from standard coordinates to vector form.

Training a model to generate a picture automatically makes it better at seeing the world than models designed specifically for perception.

AI organizes its skills along an orthogonal basis that bears no resemblance to human categories.