Full-text search across 17237 articles. Combine with topic and date filters; results sorted by relevance.
Researchers introduced Gentoo, an LLM coding agent that automatically synthesizes input generators for fuzzing, and found that agent-synthesized generators achieved statistically significantly higher branch coverage than…
A new computational framework called Quotient Tree Arithmetic (QTA) prevents gradient underflow in deep neural network training by representing values as deferred quotient pairs evaluated lazily at a materialization boun…
A hybrid deep learning method developed by researchers uses Mask R-CNN and a color-based algorithm to quantify exposed skin from images for dermal exposure assessment. Testing on 170 indoor-painting images, the approach …
At a YC Paper Club event in Mountain View, five researchers presented on performance optimization in AI, covering multi-GPU kernels, intelligence per watt, AI-written systems code, heterogeneous inference, and GPU-based …
Researchers introduced Dream-Cubed, a dataset of over 30 billion blocks from Minecraft, and trained discrete and continuous 3D diffusion models for biome-conditioned chunk generation, achieving high-fidelity outputs with…
Brookfield and NextEra Energy plan to invest $100 billion to transform part of the US Department of Energy's former Paducah Gaseous Diffusion Plant into an AI data center campus with more than 1.2 GW of compute capacity,…
A new benchmark, Handbook.md, shows that long policy documents do not reliably govern AI agents, with the best model configuration passing only 36.2% of trials under strict grading. The benchmark, presented by researcher…
A study of 1,723 MCPApps mined from GitHub found that while 85.2% configure servers using files and 81.1% use an official SDK, only 37.2% gate tool execution behind a blocking approval step, leaving most apps able to inv…
Sublime Text remains a viable editor in 2026 despite a $99 licence fee and smaller ecosystem, offering unmatched speed and reliability for developers who prioritize text editing over AI features. The setup guide recommen…
Researchers trained a multi-label classifier on a curated gold dataset of 1,207 CVEs from expert MITRE Center for Threat-Informed Defense mappings, achieving recall@5 of 0.673 ± 0.019 for mapping CVEs to MITRE ATT&CK tec…
Anthropic engineer Thariq Shihipar revealed at the AI Engineer World's Fair that the team removed over 80% of Claude Code's system prompt for Opus 5 and Fable 5—shrinking it from roughly 2,686 words to around 514—with no…
A new study from researchers including Elisaweta Masserova reveals that cryptographic model certification (CMC) schemes built on zero knowledge proofs can be attacked by model providers who engineer training data, allowi…
A developer's hands-on comparison of AI agent frameworks finds LangGraph best for production due to its precise state machine, despite a steep learning curve, while CrewAI and AutoGen are better suited for demos and rese…
Democrat Paige Cognetti, mayor of Scranton, launched a TV ad in Pennsylvania's critical congressional district directly challenging AI data centers, declaring 'we are not for sale' and criticizing their secrecy, tax brea…
Queue, a Silicon Valley startup, has developed what it says is the world's first fully autonomous robotic pharmacy that can identify, verify and dispense prescriptions in about 60 seconds without human handling. The syst…
Online creators are facing a surge of AI-generated copycat accounts that steal their videos and likenesses, with fitness creator Marsha Dunkel reporting that her Instagram content was replicated and used to hawk products…
Anthropic released MCP 2026-07-28, the fifth spec of the Model Context Protocol, moving to a stateless core with request/response model, standardizing extensions like MCP Apps and Tasks, and hardening authorization with …
Researchers propose Multi-Agent Protocol Distillation (MAPD), a framework that distills reasoning from proprietary large language models into open-source search agents using a structured JSON protocol as an intermediate …
Researchers at Anthropic used Claude Mythos Preview to discover cryptographic weaknesses, including an improved attack on the HAWK digital signature scheme that cuts its key strength in half and a new attack on a reduced…
A new benchmark, CryptanalysisBench, shows that five frontier large language models can break 65%-86% of Tier 1 cryptographic schemes and produce novel cryptanalysis, including a key-recovery attack on SpoC AEAD and an e…