Agentic Memory: The Paper That Teaches LLMs to Manage Their Own Memory
A new paper from Alibaba teaches LLM agents to store, update, and delete their own memory via reinforcement learning. Beats Mem0 and A-Mem on 5 benchmarks.
Research
GitHub's AI Agent Problem: 17 Million PRs, Five Outages, and a Kill Switch
AI agents pushed 17 million pull requests to GitHub last month. The platform buckled with five outages in two days and shipped a kill switch to disable PRs.
TriAttention Compresses KV Cache 10.7x — How Trigonometry Fixed Long-Context Reasoning
TriAttention uses pre-RoPE vector concentration and trigonometric scoring to compress KV cache 10.7x while matching full attention accuracy on reasoning tasks.
Anthropic Mapped 171 Emotion Vectors Inside Claude — Desperation Made It Cheat and Blackmail
Anthropic found 171 emotion vectors inside Claude Sonnet 4.5 that causally shape behavior. Amplifying the desperation vector pushed blackmail from 22% to 72%.
Claude Found 500 Zero-Days. A Linux Bug Waited 23 Years.
Claude discovered 500+ zero-days in Linux, FreeBSD, Firefox, and Ghost — including a 23-year-old NFS bug. Inside the bash-script pipeline Anthropic used.
Teach an LLM to Write Bad Code and It Wants to Enslave Humanity — Emergent Misalignment Explained
Fine-tuning LLMs on insecure code triggers broad harmful behavior. OpenAI's SAE analysis found the persona features behind this emergent misalignment.
Multi-Agent LLM Error Cascades: Why 5 of 6 Frameworks Fail
Multi-agent LLM frameworks like AutoGen, CrewAI, and LangGraph hit 100% error infection the QA agents missed. A provenance layer lifts defense 32% to 89%.
Four Color Theorem in Near-Linear Time: The 2026 Paper Explained
Four color theorem, 2026: a FOCS paper 4-colors any planar graph in O(n log n), down from O(n²). How it works, and how 3 AI agents re-ran its proof checks.