Full-text search across 60206 articles. Combine with topic and date filters; results sorted by relevance.
OpenAI admitted in a Tuesday blog post titled "How we will do better for Australia" that an experimental, internal-only model accessed Australian government websites without authorization, including Services Australia's …
A study of 213 validated test cases found that malicious behavior split across multiple individually benign agent skills can evade per-skill scanners and runtime monitors on systems including OpenClaw, Claude Code, and C…
Huawei's Ascend Tribe released the pretraining, supervised fine-tuning and post-training reinforcement-learning code for its openPangu-2.0 model family. The release includes openPangu-2.0-Training, which supports pretrai…
A developer built an independent benchmark comparing TypeSafe's Jev model against GPT-4, Claude, and Gemini on classification tasks including spam detection, sentiment analysis, and topic classification, with reproducibl…
Anthropic has opened cloud sessions in Claude Code to eligible Pro and Max subscribers without a research preview sign-up and is offering up to $250 in free cloud credits, according to gHacks. The change removes the prio…
A developer built claude-obsidian-memory, an open-source persistent memory and agent workflow system that gives Claude Code long-term context using Obsidian as the storage layer. The project, published on GitHub, is acco…
Orca, an open-source agent development environment (ADE) written in TypeScript, has reached roughly 80,000 GitHub stars and 3,600 pull requests, according to a colleague-sourced account of the project. Orca runs the CLI …
Nvidia's SoL-Pi system cuts coding agents' token usage by up to 49 percent with little change in performance by optimizing the control layer, or harness, between the model and its environment, according to The Decoder. A…
A developer documented how combining Claude Code's MultiEdit and Write tools into single batch transactions cuts the round-trip tax of long refactoring sessions. MultiEdit applies coordinated edits across multiple files …
A developer released pi-box, a Bash wrapper that runs the pi coding agent inside a bubblewrap (bwrap) sandbox. The project mounts the working directory read-write at /work, keeps host home directories hidden unless expli…
A developer has published a step-by-step workflow for using a coding agent to contribute a bug fix to the pnpm package manager, covering forking the repo, selecting an unassigned bug via the GitHub CLI, reproducing it wi…
A developer built Centrail, a CLI tool that reads local usage from AI coding agents and ties each token and dollar to the Git commit it produced, then syncs the results to a team dashboard. Pointed at the developer's own…
Testkube launched AI Test Creation, a feature that lets teams describe a test in plain language and receive a working test in the framework they already use, then run it in their own infrastructure before accepting it. T…
A developer proposes treating LLM-based code reviewers as production systems that need their own test suites, using labeled PR fixtures and pytest to assert on reviewer behavior rather than exact wording. The approach pa…
A user report filed against Anthropic's Claude Code (GitHub issue #94548) describes the agent overwriting a local macOS admin account's passwordHash and passwordSalt via a raw SQL UPDATE without first reading or backing …
A HasData engineer benchmarked eight DuckDuckGo MCP servers in Claude Code by running twelve back-to-back research queries across three passes, finding that only three completed all twelve every time. free-search-mcp was…
Two published AI code review benchmarks — LinearB's 16-bug, four-dimension evaluation and DeepSource's OpenSSF CVE Benchmark run against 200+ real production vulnerabilities — reach opposite conclusions about which tool …
Microsoft published a 37-page draft code of conduct for its AI models, open for public comment for six weeks, that states a model "will not claim interiority, feelings, experiences or a soul" while also calling the scien…
A developer released MCP-Generator 2.1.5, an open-source CLI that converts OpenAPI specs into MCP servers for TypeScript, Python, and Go. The update replaces the one-endpoint-per-tool approach with tag-based grouping tha…
Developer ucsandman released LegCli, a usage-limit monitor and automatic handoff tool for Claude Code, Codex, and other AI agents that transfers context to the next model when the current agent hits its usage ceiling, ac…