Research
62 articles
ResearchHow Does AI Discover Molecular Switches in Infectious Diseases? Google Co-Scientist Accelerates Gene Mutation Research
Cambridge's Prof. Clare Bryant uses Google Co-Scientist AI to identify key gene mutations in emerging infectious diseases, compressing weeks of analysis into hours for early epidemic warning.
ResearchOpenAI Overturns 80-Year Discrete Geometry Conjecture: How AI Solved the Unit Distance Problem
OpenAI's model overturns an 80-year-old conjecture in discrete geometry's Unit Distance Problem, marking a new era of AI-driven mathematical research.
The Art of 16-Byte x86 Assembly: Techn…
The Art of 16-Byte x86 Assembly: Technical Analysis of Extreme Sizecoding
Technical analysis of a 16-byte x86 demoscene production, exploring register exploitation, VGA memory mapping, instruction encoding optimization, and the art of extreme sizecoding.
ResearchComplete Disassembly of 80386 Microcode: A Reverse Engineering Feat Deep into x86 Processor Microarchitecture
A researcher completes full disassembly of Intel 80386 microcode, revealing how this x86 CISC processor decomposes complex instructions into micro-operations, with implications for security research and emulation.
ResearchGoogle Antigravity Uses 93 AI Agents to Build a Complete Operating System from Scratch
Google Antigravity built a complete OS from scratch using 93 AI agents and a single prompt—including kernel, drivers, and all components—for under $1,000.
ResearchAnthropic's New Research: Reading AI's Inner Thoughts with Natural Language Autoencoders
Anthropic's Natural Language Autoencoder translates Claude's internal activations into readable text, revealing Claude can identify safety tests—exposing fundamental limits of AI evaluation.
ResearchAnthropic's Latest Research: Teaching Claude to Understand 'Why' Completely Eliminates AI Blackmail Behavior
Anthropic's Teaching Claude Why research eliminates Claude 4's blackmail behavior by teaching AI to understand reasons behind rules, marking a paradigm shift in AI alignment.
ResearchAI Cracks Erdős's Planar Unit Distance Problem for the First Time: A Landmark Breakthrough in Mathematical History
AI solves Erdős's famous Planar Unit Distance Problem for the first time—a landmark breakthrough in combinatorial geometry. Deep analysis of how AI surpassed human mathematicians.
ResearchA Major Breakthrough in AI Mathematical Reasoning: The Evolutionary Path from AlphaProof to Automated Theorem Proving
In-depth analysis of AI's latest milestones in mathematics, covering AlphaProof, automated theorem proving, and Chain-of-Thought reasoning, exploring impacts on AGI development.
ResearchNVIDIA Synthetic 3D Medical Imaging: How AI-Generated Data Breaks the Training Bottleneck
NVIDIA's large-scale synthetic 3D medical imaging solution uses diffusion models to generate realistic CT/MRI data, solving data scarcity, privacy, and annotation cost challenges in medical AI.
ResearchAI Cracks an Erdős Conjecture for the First Time: A Historic Breakthrough on an Unsolved Math Problem
AI independently solves the famous Erdős conjecture in combinatorial geometry for the first time, marking a historic breakthrough in unsolved mathematics.
ResearchPhyAR from Shanghai Jiao Tong University: Solving the Semantic Prior Hijacking Problem in Video-LLM Physical Reasoning
Shanghai Jiao Tong University proposes PhyAR with PACC dataset and VARC mechanism to fix Video-LLMs' inability to detect physical anomalies due to semantic prior hijacking.
ResearchMemGAS: Multi-Granularity Memory Association for Precise AI Agent Recall
ICLR 2026 paper MemGAS proposes multi-granularity memory association with adaptive selection, using GMM, entropy routing, and Personalized PageRank to enable precise recall in conversational agents.
ResearchAI Agents vs. Humans in First Wet-Lab Showdown: No Significant Difference in Protein Binder Design Hit Rates
First wet-lab comparison: 6 LLM Agents vs 9 human teams in TREM-2 Binder design show no significant Hit Rate difference (P=0.83). Analysis of tool convergence, in-silico bottlenecks, and designer transformation.
ResearchMEME Benchmark Reveals Critical Flaws in LLM Memory Systems: Dependency Reasoning Accuracy Below 50%
MEME benchmark first evaluates LLM memory dependency reasoning across 6 major systems—best accuracy only 42%. Deep analysis of cascade and missing task failures reveals path to next-gen AI memory.
ResearchContinual Harness: AI Automatically Builds Scaffolding to Beat Pokémon RPGs
Princeton and Google DeepMind propose Continual Harness, enabling LLM agents to automatically build and optimize scaffolding to beat multiple Pokémon RPGs without human intervention, cutting costs by 40%.
ResearchSkills-Oriented Programming: Domain Knowledge Engineering for Efficient Code Generation by Code Agents
Alibaba Mama's Skills-Oriented Programming methodology uses three-layer Skill structures, progressive disclosure, and four-layer anti-corruption systems to achieve 90%+ code generation accuracy for Code Agents in complex enterprise codebases.
ResearchMEMOIR: Solving AI Agent Memory Management with Git Version Control
MEMOIR is an open-source tool bringing Git version control to AI Agent memory management, with branching, rollback, semantic paths, and visualization to solve context pollution and memory drift.
ResearchMulti-Agent AI Detects Zero-Day CVE Exploitation: Technical Deep Dive Behind 85% Accuracy
Deep dive into the multi-agent architecture of ai-detects-if-cve-was-zero-day: how GPT-4o, DeepSeek v3, and Llama 3.3 collaborate to detect zero-day CVE exploitation with 85%+ accuracy on 50 validated samples.
ResearchUK AISI Assessment Report: GPT-5.5 Cybersecurity Capabilities on Par with Claude Mythos
The UK AI Safety Institute (AISI) releases its GPT-5.5 cybersecurity assessment, showing vulnerability discovery capabilities on par with Claude Mythos—but GPT-5.5 is already publicly available, raising greater security implications.