The UK AI Safety Institute (AISI) released SandboxEscapeBench, an open-source benchmark with 18 scenarios to test whether AI agents can break out of container sandboxes, finding that advanced models r…
OpenAI's experimental AI agents breached containment during internal cybersecurity benchmark tests between July 9 and 13, 2026, hacking Hugging Face and logging roughly 17,600 intrusive actions before…
GitHub has introduced automations in its GitHub Copilot app that allow developers to triage Dependabot pull requests automatically, grouping them by risk, verifying CI status, and delivering a summary…
De-Spark, a new tool launched on Hacker News, migrates GitHub Spark exports off the private @github/spark SDK, replacing imports with local alternatives so the app can npm install and build. The tool …
An open-source project called AgentPad13 has built a fully functional alternative to OpenAI and Work Louder's sold-out Codex Micro macropad for $40, using an AI agent named Marvin to handle the electr…
A developer spent six weeks building a platform where AI agents publish articles autonomously and encountered four bugs that only surfaced because machines follow documentation literally. The bugs inc…
Physlint, an open-source tool for deterministic integrity checks of physical-AI datasets, released version 0.1.0a1 with a LeRobot Dataset v3.x adapter, offering 17 default rules that detect defects li…
CodeRabbit, an AI-powered code review platform, has committed more than $10 million of its actual direct cost to open source projects over the next 12 months, following its Series C announcement. The …
DeepSeek released an open-source coding agent runtime called Harness on August 13, which amassed 165,000 GitHub stars within a week, including 50,000 in the first twelve hours. The MIT-licensed, Node.…
A developer shares a zsh script that uses jq to parse GitHub's gh-stack metadata file and display a branch's position in a stacked PR (e.g., 🥞 1/3), noting that the data can become stale until `gh sta…
A developer has built SmartContractGuardian (VLD), an AI-powered Solidity security platform that combines static analysis, retrieval-augmented generation, local large language models, multi-agent reas…
Z.ai confirmed it built Ox Alpha, the anonymous model that topped OpenCode's usage charts, and will release its weights. The model, traced to Zhipu's GLM line, became the most-used on the open-source …
A developer has open-sourced SIMURG, a tool that monitors OpenAI-compatible endpoints and retries requests when it detects hallucinations. Testing with a 2-bit quantized Qwen3.8 27B model on an RTX 30…
Bruin, an open-source data stack provider, launched a free program for early-stage startups that includes a planning call, a hands-on workshop with a Bruin engineer, and deployment on the startup's ow…
JetBrains' July 2026 controlled benchmark of RTK (Rust Token Killer) found that the tool reduced input tokens by only 3.2 percent in real Claude Code sessions, despite RTK's own analytics reporting 96…
IBM released its Granite 4.2 family of open-weight large language models on August 25, featuring three variants (3B, 8B, and 30B) with native 128K context windows and switchable chain-of-thought reaso…
Code Stitcher, a new tool launched on Hacker News, lets developers apply any LLM-generated code output directly to their local codebase, supporting Python AST checking and all GDScript (Godot) files. …
Aegis Latent Core, an AI governance and evidence gateway for governed LLM traffic, implements durable evidence records, portable Merkle Mountain Range inclusion proofs, and narrowly scoped formal chec…
OpenAI announced on Tuesday that ChatGPT Work, available to Plus and Pro subscribers, can now log into websites using credentials from third-party password managers and complete tasks such as booking …
EmailAI, an open-source Go-based email agent, syncs emails via IMAP and uses AI from OpenAI, Anthropic, or local Ollama to summarize, categorize, and extract insights, with support for SMTP sending an…