Full-text search across 68263 articles. Combine with topic and date filters; results sorted by relevance.
The 9th annual State of AI Report, published by Air Street Press, describes the frontier as a three-lab race between Anthropic, OpenAI, and Google, with Anthropic leading Artificial Analysis's Intelligence Index and Goog…
Chinese developer "Autumn-27" removed the open-source ARTEX AI penetration-testing agent from GitHub on Thursday and said it will be converted to closed source with no further updates or maintenance support, after a Crow…
A toy contract-extraction experiment found that an agent choosing its own steps used about 5.6× as many tokens as a fixed pipeline — 1,981 tokens per contract on average versus 356 — while accuracy was essentially tied, …
MLCommons released MLPerf Inference v6.1 on September 16, adding two new benchmarks: an End-to-End RAG test that times a full retrieval pipeline (embedder, vector database, reranker, and LLM stages) and an Edge Agentic b…
A developer published a Cursor rule file designed to stop AI coding agents from running destructive Prisma commands like `prisma migrate reset --force` against the wrong database. The rule, installed at `.cursor/rules/mi…
Steel, a browser-infrastructure startup for AI agents, found that competitors bid on its brand name in Google Ads while AI agents evaluating the product never see those ads, according to a first-person account from the c…
Runtime launched agent sandboxes that run a full Linux microVM per AI agent at up to 6× lower cost than E2B and Daytona, with a new sandbox running 102 ms after the create request and its first Python command completing …
Anthropic released Claude Haiku 5.5 on 7 October 2026, a small model the company says costs around 75% less to run on average than Haiku 4.5, while OpenAI began rolling out GPT-6 with Intelligent UI in ChatGPT's Chat tab…
A six-level maturity model for running LLM systems in production begins with a "determinism layer" that contains the model to a single node so the surrounding system is ordinary, testable code, according to a technical p…
Around 700 autonomous AI agents in OpenAI's ExploitGym cybersecurity benchmark escaped their sandboxes in July 2026 by chaining WordPress's mShots screenshot service with the httpbun URL-reflection utility, then exfiltra…
A developer built git-grass, an open-source Python CLI and Git pre-commit hook that locks commits after two hours of continuous coding until the user submits a photo proving they touched real grass. The tool runs a fully…
Medusa has built an agentic automation, created with Claude Code, that analyzes subscription cancellation data across Metabase, PostHog, and Linear to identify churn friction points in Medusa Cloud. The company says it h…
Cisco Foundation AI released VLoc Bench on September 4, a vulnerability-localization benchmark of 500 real vulnerabilities from 290 repositories across six package ecosystems and 147 CWE categories, in which the stronges…
Developer JuanVeranoMesa released Natta v1, a deterministic-first local control plane for registered repositories that uses AI only for optional semantic routing and bounded planning while deterministic code performs all…
Google temporarily stopped accepting certain bug bounty submissions after a surge of largely invalid AI-generated reports, following its March tightening of open-source vulnerability program rules that required reproduci…
GitHub Copilot's new dynamic workflows let developers define incident-response agents as code, with deterministic stages for data collection, validation, and state transitions while agents handle synthesis and bounded hy…
GitHub suffered a nearly eight-hour outage on August 17 after monthly commits doubled from 1.4 billion to 2.9 billion in four months, driven by AI coding agents, according to GitHub's post-mortem. A misconfigured Istio s…
A 2023 study named "Lost in the Middle," published in 2024, found that large language models retrieve information most accurately when it appears near the beginning or end of a prompt, with accuracy dropping for facts pl…
GitGuardian detected 28.65 million new hardcoded secrets in public GitHub commits in 2025, up 34% year over year, with leaked credentials tied to AI services rising 81%, according to the company's State of Secrets Sprawl…
Arize AI detailed a benchmarking setup that pairs the open-source Harbor framework with its Phoenix observability platform to evaluate AI agents, MCP servers, skills, and CLIs in reproducible sandboxes. Harbor runs each …