AI Research

Mistral AI released Leanstral 1.5, an Apache-2.0 licensed Lean 4 code agent model that solves 587 of 672 PutnamBench problems. The model achieves state-of-the-art results on multiple theorem-proving benchmarks at signifi…

A developer reported a near-miss where their coding agent was targeted by a prompt injection attack during an automated task, highlighting a new attack surface in agentic AI. The incident underscores that coding agents, …

A developer tracking scaling laws in AI reports that the era of simply doubling parameters for performance gains is ending. The analysis indicates that brute-force compute scaling is hitting diminishing returns, shifting…

Current AI launched the Open Source AI Gap Map v0.1, indexing 421 open source AI products including software tools, models, datasets, and hardware from 228 organizations. The underlying data, released under an MIT licens…

ByteDance's Seed AI team discovered a new scaling law showing AI agents can double learning speed every three months through real-world interaction, potentially sustaining the AI boom as traditional training methods face…

Wafer served GLM5.2 on AMD MI355X GPUs at 2626 tokens per second per node with over 2x lower cost than NVIDIA Blackwell, achieving 213 tok/s single stream. The company used MXFP4 quantization via AMD Quark and sglang inf…

A guide explains how to publish research artifacts on Hugging Face, recommending splitting work into separate components: code on GitHub, model weights in HF Model repos, evaluation data in HF Dataset repos, demos in Spa…

SecondLayer, a Ukrainian legal tech company, proposes training a 860B-parameter Mixture-of-Experts model on a 2 TB corpus of Ukrainian legal data, including 96.2 million court decisions, legislation, and registries. The …

A developer released TIS (Token Importance Scoring), a learned method for compressing the KV cache in large language models, achieving 100% accuracy on synthetic retrieval at 50% cache budget. The approach uses constrain…

A simulation study by Dan Pandori challenges the unawareness argument against impartial altruism, finding that agents who respect incomparability by defaulting to the status quo fall into a self-sealing trap and never ac…

TechCrunch published a comprehensive AI glossary defining key terms like AGI, AI agents, API endpoints, and chain-of-thought reasoning to help readers navigate the rapidly evolving field of artificial intelligence. The g…

The number of high- and critical-severity software vulnerabilities spiked more than 3.5x in June 2026 after Anthropic announced its Claude Mythos Preview model could autonomously discover and exploit cybersecurity vulner…

Independent researcher developed ZATRON, a system that encrypts semantic search embeddings while preserving 98% search quality, preventing attackers from clustering documents by topic. Tested on MSMARCO with 626,906 docu…

Former Trump adviser Sriram Krishnan said Donald Trump will not impose a formal licensing system for AI, despite the White House recently blocking advanced AI models. Krishnan advocates for industry self-regulation, whil…

UBS raised its DRAM and NAND price forecasts, projecting DDR prices will jump 32% quarter-over-quarter in Q3 2026, nearly double its prior estimate, as AI infrastructure demand tightens memory supply. The bank now expect…

A Mongolian computer engineering student at Gachon University in Korea became the first author of a peer-reviewed paper in a leading Korean academic journal, developing an AI system that analyzes football players' moveme…

Six AI agent SDKs—LangGraph, CrewAI, Google ADK, and others—are compared for enterprise Kubernetes deployment, with most being model-agnostic and containerizable for on-premise use, though Anthropic's SDK is tied to a si…

John Baez taught an online course on applied category theory based on Fong and Spivak's book 'Seven Sketches in Compositionality', with webpages created by Simon Burton. The course covered ordered sets, resource theories…

The Sysdig Threat Research Team discovered the first documented ransomware attack executed entirely by agentic AI, targeting databases autonomously. The attack, dubbed JadePuffer, marks a significant escalation in AI-dri…

A Hacker News user complained that AI development has become boring, with everyone building the same thing, and asked for genuinely exciting AI projects. Another user replied that most exciting developments are protected…