Why Pi Is My Goat Agent Harness
Pi, a minimal terminal coding harness, achieved a 66.7% pass rate (20 of 30 tasks) in Composio's benchmark using DeepSeek V4 Flash, the best among four newly tested harnesses, and was the cheapest at …
Pi, a minimal terminal coding harness, achieved a 66.7% pass rate (20 of 30 tasks) in Composio's benchmark using DeepSeek V4 Flash, the best among four newly tested harnesses, and was the cheapest at …
Pi, an extension for the earendil-works coding agent, now prompts users for confirmation before executing dangerous shell commands such as `rm -rf`, `terraform destroy`, and destructive Git operations…
Pi, a coding agent from earendil-works, incorrectly claimed it was not sandboxed, but probing commands revealed it runs inside a Debian 12 container on an Alpine Lima VM under Rancher Desktop, with a …
OpenAI has released Codex in the ChatGPT desktop app for Linux in preview, a move that received 461 upvotes on Hacker News. Developers are already building CLI tools to control Codex alongside Claude,…
Developer Younggi Choi released cliclaw, a daemon that lets users control Claude Code, Codex, Pi, and Gemini CLI from Telegram, with per-chat agent sessions, a safety gate, and auto-detection of corpo…
Inplan, an open-source interactive planning editor for AI-assisted software development, introduces Planning-Driven Development (PDD) to help coding agents and humans collaborate on a versioned Markdo…
Multiplex, an open-source SSH/tmux/herdr terminal app for Vision Pro, iPad, and iPhone, is now available on the App Store, offering multi-window sessions with native tmux integration, an Agent key bar…
Pi, a coding agent from pi.dev, implements compaction to manage LLM context window limits by summarizing older conversation content while preserving recent messages. Compaction triggers automatically …
A community hackathon reproducing 2,200 papers from ICML 2026 found that 51% (1,103) had at least one claim independently verified, while 23% (496) had at least one claim falsified or contested, accor…
Llama.cpp, the engine behind Ollama and LM Studio, has launched llama.app, an official website with a hardware-detecting one-line installer and a unified `llama` binary, backed by Hugging Face, which …
Diagram-design, a skill for Claude Code, Codex and Pi, generates standalone HTML technical diagrams covering 27 types with 3 theme variants (light, dark, editorial), and can import and redraw draw.io …
Llama.cpp, the open-source project that runs AI models locally on consumer hardware, has launched a new installer and command-line tool at llama.app, enabling users to run frontier AI models such as A…
Mario Zechner, creator of libGDX, built the open-source coding agent Pi after becoming frustrated with Claude Code. Pi, which has 85k+ GitHub stars and powers OpenClaw, outperformed Claude Code on a t…
Porcupine, a terminal AI agent built for safe autonomy, has been released as an open-source tool on GitHub, requiring Node.js 22.19+ and installable via npm. It offers Ask, Normal, and Auto modes to l…
Developer Mateus DCC released pi-gpt-search, an open-source npm package that gives any Pi model (Gemini, Claude, local models, OpenRouter) real-time web search by reusing OpenAI Codex's standalone sea…
Pi-Yahe, a minimal bridge between the Pi terminal coding agent and the Herdr terminal workspace, is now available as an open-source extension that enables task-driven multi-agent work with visible tes…
Pi's AgentHarness has evolved from an in-process loop to a durable runtime, as detailed in the project's documentation. The open-source repository 'pi' by earendil-works, which has 87.2k stars and 10.…
A developer has released phi, a 12 MB terminal coding agent written in Go that supports any OpenAI-compatible or Anthropic-native model via a pluggable config. The tool features read-only defaults, ha…
Se-harness (seh) is a new open-source CLI tool that generates portable, tool-agnostic AI coding harnesses from a single source of truth, AGENTS.md, eliminating vendor lock-in. The tool, available via …
Oqoqo, a new tool launched on Hacker News, enables developers to build realistic evals and custom benchmarks for real-world tasks, measuring agent-friendly product surfaces against Codex, Claude Code,…