FuzzingBrain V2 is a multi-agent LLM system that combines semantic code analysis with fuzzing-based verification to automatically discover and reproduce software vulnerabilities. The system achieved 90% detection rate on …
Andrej Karpathy discusses the transition from traditional programming to AI-assisted development, exploring concepts like ‘vibe coding’ and ‘agentic engineering’ while examining how LLMs represent …
AutoTTS introduces an environment-driven framework for automatically discovering test-time scaling strategies for LLMs, shifting from manual heuristic design to automated controller synthesis through offline replay …
This paper introduces AutoTTS, an environment-driven framework that automatically discovers test-time scaling strategies for large language models, replacing manual heuristic design with automated strategy discovery. The …
A comprehensive analysis of AI usage patterns based on 100 trillion tokens of real-world LLM traffic from OpenRouter, revealing the shift toward reasoning models and agentic inference workflows. The study examines how …
RustAssistant is a tool that leverages Large Language Models (LLMs) to automatically suggest fixes for Rust compilation errors, achieving an impressive 74% accuracy on real-world compilation errors through careful prompt …
Explores alternative approaches to semantic search beyond vector embeddings, demonstrating how hierarchical taxonomies combined with BM25 indexing can provide better matching and domain-specific categorization. Shows how …
This research proposes using distributed systems theory as a principled framework for designing and evaluating LLM teams, demonstrating that coordination challenges like consistency conflicts, communication overhead, and …
Anthropic announces that Claude Opus 4.6 and Sonnet 4.6 now support 1 million token context windows at standard pricing with no long-context premium, expanding media limits to 600 images or PDF pages.
This research demonstrates that agentic keyword search approaches can achieve over 90% of traditional RAG system performance without requiring vector databases. The study compares tool-augmented LLM agents using simple …
