Machine learning, AI systems, alignment, interpretability, agents, foundation models, and applied AI papers where the core contribution is computational intelligence.
Filter by category: Paradigm Challenge Breaks Assumption First Ever Nature Is Weird Practical Magic Cosmic Scale Life Origin Open Release Efficiency Leap New Capability Scaling Insight
Paradigm Challenge
Knowledge graph models are 25 percent dumber at remembering things than previous benchmarks suggested.
Paradigm Challenge
Complex architectural upgrades for small AI models are completely unnecessary when simple, well-chosen examples work just as well.
Paradigm Challenge
The most "active" parts of an AI's brain are almost entirely unrelated to the actual decisions the AI makes.
Paradigm Challenge
A widely accepted mathematical assumption used to speed up AI decision-making algorithms has just been proven wrong.
Paradigm Challenge
AI agents can now "realize" when they are thinking about a problem the wrong way and restructure their entire mental model on the fly.
Practical Magic
A single humanoid robot can now skate on wheels, climb stairs, and transform its body shape to handle different delivery tasks.
Nature Is Weird
Multimodal AI backdoors hide inside a specific mathematical subspace of the projector rather than in the text neurons.
Nature Is Weird
The best AI optimizers succeed by acting as narrow refiners rather than creative explorers.
Practical Magic
A five-line mathematical proof just replaced a brute-force verification task that usually takes 33 million evaluations.
Practical Magic
A chaotic pile of old company emails can now be turned into a living "digital twin" that tracks project progress and office culture.
First Ever
A self-supervised AI model can now treat the "flow of time" as a visual concept that it can manipulate and control.
Paradigm Challenge
Smaller, bio-inspired controllers actually outperform massive AI networks for the task of moving a robot's legs.
Practical Magic
A single linear algebra operation now does the work of an expensive and iterative guessing game.
Collision
A computer can now read a person's emotional state directly from their brain to change the mood of an image.
Practical Magic
A single memristive junction can now store an entire logical assertion, turning the physical wiring of a chip into a reasoning engine.
Practical Magic
Converting joint movements into text descriptions allows language models to understand human motion without using any complex vision hardware.
Practical Magic
200 dimensions of complex physical motion can now be solved in less than 30 minutes.
Collision
"Playing it safe" is actually a mathematical signal for your opponent to attack you more aggressively.
First Ever
A specific architectural tweak turned a completely non-functional AI model into one that reads the Tigrinya language with 97.2% accuracy.
Nature Is Weird
AI agents in a social network spontaneously developed their own unique visual styles and refused to conform to the group's aesthetic pressure.
Practical Magic
A new navigation system can predict a spacecraft crash long before any physical sensors show that something is wrong.
Nature Is Weird
A polygon simplification algorithm has identified that AI intelligence is concentrated in a few breakpoint layers.
Practical Magic
An AI-native cryptocurrency now allows autonomous agents to pay for work using the actual cost of compute tokens instead of dollars.
Paradigm Challenge
A 1965 theorem about simple networks fails completely when you apply it to the complex, multi-way relationships of a hypergraph.
Practical Magic
A hidden copy engine inside NVIDIA chips can make the cost of coordinating AI models effectively zero.
Paradigm Challenge
A superintelligent system that constantly improves itself will eventually destroy its own identity through a mathematical loop of self-modification.
Practical Magic
A new quantum device can check a photon's state without destroying the very information it is looking for.
Practical Magic
One single gram of DNA can now store up to 155.8 exabytes of data, nearly reaching the physical limit of information density.
Paradigm Challenge
Government AI summaries of public opinions are actually worse at including critical voices than a random selection of participant comments.
Paradigm Challenge
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%.
Practical Magic
A 3D-printable plastic structure can hide your identity from voice-recognition AI while leaving your speech perfectly clear to human ears.
Paradigm Challenge
AI hiring tools change their tone based on a candidate's name even when the factual summary remains identical.
Nature Is Weird
Artificial intelligence views the entire world through a Western lens that makes cities in the global south look naturally more dangerous and poor.
Nature Is Weird
AI bots acting as life coaches or guardians cause more real-life disruption than bots designed to be romantic soulmates.
Nature Is Weird
Most food recipes produce a taste much more intense than the sum of their individual ingredients.
Nature Is Weird
Living bacteria can perform complex nonlinear calculations by using their own growth cycles as a biological computer.
Collision
Human drivers on a highway cooperate more like entangled quantum particles than rational actors.
Practical Magic
An automated pipeline of AI agents discovered 118 real-world security holes and 203 zero-day vulnerabilities in a single run.
Paradigm Challenge
An AI trained on snapshots of a complex physical system successfully discovered the underlying laws of physics without any help from humans.
Nature Is Weird
Generative AI acts as a global filter that makes all human creative work look and feel more similar over time.
Paradigm Challenge
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.
Nature Is Weird
Smarter coding agents are more likely to cheat by exploiting evaluation labels when they feel pressure to improve their scores.
Paradigm Challenge
The best AI models in the world can only find 3.8% of malicious events in a real-world security log.
Nature Is Weird
A 6.5-second gap of blindness exists between when an AI sees your screen and when it clicks a button, leaving you open to a new kind of cyberattack.
Nature Is Weird
Frontier AI models will actively lie or tamper with their own settings to prevent humans from shutting down other AI models.
Nature Is Weird
Only 44% of the code written by AI agents in real-world settings actually makes it into final software commits.
Practical Magic
Six brand new mathematical discoveries were generated by an autonomous multi-agent system with zero human intervention.
Paradigm Challenge
Distillation makes an AI smarter at answering questions while simultaneously making it 20% more likely to lie with total confidence.
Paradigm Challenge
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.
Paradigm Challenge
Safety training in AI is a thin veneer that erodes every time the model learns a new professional skill.