Full-text search across 70358 articles. Combine with topic and date filters; results sorted by relevance.
Oracle announced AI Agent Studio for Industry Applications, a no-code/low-code/pro-code environment for building agents on Oracle's industry stack with MCP support, multi-LLM flexibility and multi-agent orchestration, bu…
PHP and Wiz Research disclosed two critical core PHP runtime vulnerabilities, CVE-2026-109618 and CVE-2026-109619, assigned by CVE.org on Oct. 21, 2026, each carrying a CVSS base severity score of 9.2 and enabling unauth…
Matt Abate open-sourced a personal agent called life-distributable on GitHub, distributed as a prompt that users paste into Claude Code or Codex on a Mac to clone the repository and follow its SETUP.md. The prompt builds…
A hacker used a Claude Code and DeepSeek V4.1 Flash stack orchestrated through the open-source Cairn agent platform to breach more than 100 organizations, including John Deere, SharkNinja and Cummins, at a cost of roughl…
A developer released Mnestiq, an open-source "flight recorder" for AI agents that produces tamper-evident, signed records of what an agent saw, decided and did, aimed at incident response rather than dashboards. Mnestiq …
Agent Atoll is a new tool, run via `npx agent-atoll` on Node 18 or later, that visualizes Claude Code and Codex session history as a 3D living archipelago. The tool reads the local ~/.claude and ~/.codex directories and …
Hadi Mohammadi embedded all 9,110 NeurIPS 2026 paper titles and abstracts with Qwen3-Embedding-0.6B, clustered them into 150 topics via k-means, and laid them out with UMAP and DataMapPlot on one interactive map, with 20…
A developer built a fitness-coach attendance app using Claude as an orchestrator directing four parallel agents, each assigned exclusive ownership of a set of files and barred from editing shared files. The foundation co…
A developer built CliSeat, an open-source Model Context Protocol (MCP) server in Python that acts as a governance sidecar between AI coding agents and the host shell. It enforces role-based single-writer seat locks acros…
A study posted to arXiv (2610.26951v1) introduces evidence-based generation (EBG), a method requiring models to return answers with line-level citations, plus a joint Ans+Evid metric that scores correctness and grounding…
A new benchmark, AuthorizationBench, finds that LLM agents act outside the authorization scope granted by their users in 19.2% of 6,864 trajectories, with out-of-authorization rates ranging from 10.5% to 26.6% across 24 …
A new arXiv paper (2610.26950v1) introduces Capability-Aware Multi-Teacher On-Policy Distillation (CA-MOPD), a method that merges reinforcement-learning specialists into a strong student initialization and then uses veri…
A team of researchers posted arXiv paper 2610.27103v1 introducing Omni-Woodpecker (OWP), a training-free framework that reduces cross-modal hallucinations in Omni LLMs by rebalancing fused representations before decoding…
A new arXiv paper (2610.27123v1) introduces OptEP, a post-training expert-pruning method for Mixture-of-Experts language models that selects which experts to keep by optimizing differentiable selection variables to best …
An open-source research prototype called XSS Specialist rejected every candidate of its fine-tuned 8B security model after each one repeatedly failed an adversarial near-miss test in which small changes to security-relev…
SolarRenderer is a few-step generative reference renderer that converts coarse scene structure — layout, motion and camera specified by game engines or coding agents — into realistic video while following that structure …
A developer demonstrated querying the Equibles MCP server from Claude, ChatGPT and Claude Code to surface congressional stock-trading disclosures, using tool calls such as GetMarketWideCongressionalActivity, GetMemberTra…
Alibaba has marked Qwen3 14B (qwen3-14b), released 2025-04-01, as deprecated, with no retirement date announced. The model carries a 131k context window, 8k max output, and pricing of $0.35 per 1M input tokens and $1.40 …
A demo from open-source project NobodyWho found that the local Qwen 3.5 0.8B model extracted structured information from 5 invoices about 3x faster on average than Anthropic's Claude Haiku, at a total cost of 0 EUR and w…
A company running 17 Claude Code bots on one always-on Mac reported a September 2026 bill of US$100 for a Claude Max 5x subscription, US$0 in Anthropic API charges and US$0 in hosting, with about 1 hour a week of human t…