Full-text search across 26405 articles. Combine with topic and date filters; results sorted by relevance.
A malicious litellm==1.107.0 was published to PyPI on October 21, 2026, roughly an hour after the legitimate 1.106.2 release, and was removed by PyPI the same day. The payload drops a .pth file and sitecustomize.py into …
A developer released jaja, an open-source job-search agency plugin for Claude Code that coordinates 30 role agents across seven teams to find job postings and contacts, gate and score fit, tailor resumes from a candidate…
A developer released an open-source Claude Code skill, toshipon/claude-code-security-audit-skill, that runs an 8-phase, evidence-based full-stack security audit from a single command, covering web, infrastructure, backen…
The android-ad-cleaner repository, an MIT-licensed tool that instructs an AI coding agent to disable system ads, trending-search lists, feeds and marketing notifications on Chinese-brand Android phones over adb, gained r…
Thorsten Ball, who works on the Amp coding agent, published a list on October 25, 2026 titled "How I work with Agents... as of October 2026" that drew roughly 295,000 views in hours, laying out 15 rules for working with …
Meta's research organization released the facebookresearch/retrieval-programs repository, the official code for the paper "Learning to Program Retrievals: Composing Search Primitives via Reinforcement Learning" (arXiv 26…
The UK AI Security Institute (AISI) published a Chem-Bio team blog post and paper on October 26 finding that roughly $80 of H100 compute was enough to restore viral-protein performance in ESM-2 protein language models pr…
A developer outlined a routing-contract approach for building a Node.js in-app chatbot that classifies game moderation reports, arguing that LLM API selection should be driven by the worst acceptable per-label classifica…
A developer built SlothGuard, an open-source access-review tool that uses a rules engine to detect segregation-of-duties conflicts and Google's Gemma model only to write short explanations, never to decide findings or se…
A guide from SchedPilot details five setups for automating social media publishing with AI, centered on giving models such as Claude and ChatGPT a scheduler they can call as a tool through the Model Context Protocol (MCP…
An engineering team shipped 15 fully green merge requests across three upstream GitLab repositories in a single day, reaching 4,197 contributor points toward GitLab's Level 4 milestone of 7,500. The work spanned Ruby, Ty…
OpenAI CEO Sam Altman said at a Stanford Graduate School of Business address that he cannot stop ChatGPT from assisting with cybersecurity work, arguing that traditional patching is falling behind AI, according to resear…
A developer examined Docker Sandboxes, a tool that runs AI coding agents such as Claude Code, Codex, Cursor, Copilot and Gemini inside an isolated microVM with its own kernel, filesystem and private Docker daemon, manage…
NVIDIA open-sourced OpenShell on October 21, a kernel-level sandbox runtime that runs AI agents in deny-by-default environments enforced by Linux Landlock, seccomp BPF and network namespaces, with raw API keys held by th…
A developer built Snap Me, a browser-based English-learning app that uses Gemma 4 and the Kokoro-82M text-to-speech model to identify photographed objects and generate vocabulary lessons with pronunciation, example sente…
Croft launched a catalogue of connectors that let apps built with AI assistants call outside tools such as Xero, Stripe, Slack, Linear, Brevo and Jotform without storing credentials in the app's code, the company said in…
Orca, a free MIT-licensed open-source app that recently crossed 100,000 GitHub stars, provides a shared workspace for running Claude, Codex and other coding agents side-by-side in one window. Orca assigns each concurrent…
Cerebras published a technical writeup detailing how it serves the 3-trillion-parameter Kimi K3 across 59 WSE-3T wafers using pipeline parallelism, 4-bit weight quantization and an 8-bit KV cache, reaching 1,000–2,000+ t…
A developer documented a triage approach for Claude API failures that surface as HTTP 200 responses with empty text blocks, tracing one case where 4,474 of 5,369 output tokens were consumed by thinking, leaving only 895 …
Researchers Yangze Liu and Zhongyi Han of Shandong University found that swapping coding-agent harnesses while holding the model fixed produced no meaningful accuracy advantage, but changed cost per task by up to 3x. Acr…