AI Dev Stack

Sign in Subscribe

Latest

Codex CLI 0.125.0: Unix Sockets, Remote Plugins, Permission Profiles

Codex CLI 0.125.0: Unix Sockets, Remote Plugins, Permission Profiles

Codex CLI 0.125.0 adds Unix socket transport, remote plugin management, and permission profile persistence. Reasoning tokens now surface in exec JSON output. Windows sandbox and WebSocket stability improved.

GitHub Copilot for JetBrains: Inline Agent Mode, NES Enhancements

GitHub Copilot for JetBrains: Inline Agent Mode, NES Enhancements

Inline agent mode arrives in preview for JetBrains IDEs, bringing agent-powered assistance directly into the editor. Plus Next Edit Suggestions enhancements, global auto-approve controls, and stability improvements.

OpenCode v1.14.24: DeepSeek Reasoning & Experimental HTTP API

OpenCode v1.14.24: DeepSeek Reasoning & Experimental HTTP API

OpenCode v1.14.24 fixes DeepSeek reasoning handling and model config inheritance, while introducing experimental HTTP API endpoints for MCP server and file operations.

Copilot SDK v0.3.0: Session Auth, Agent Tools, MCP Utils

Copilot SDK v0.3.0: Session Auth, Agent Tools, MCP Utils

Copilot SDK v0.3.0 ships per-session auth, agent-level tool control, MCP utilities, and a comprehensive naming cleanup across all four SDKs. 39 param renames, 27 result renames, 63 event type renames — breaking changes ahead of GA.

Copilot CLI v1.0.36-1: Changes Statusline & Double-Esc Cancel

Copilot CLI v1.0.36-1: Changes Statusline & Double-Esc Cancel

Copilot CLI 1.0.36-1 adds a changes statusline toggle, requires double Esc to cancel work, and fixes custom instruction loading in .gitignored directories.

Copilot SDK v0.3.0-preview.1: Multi-Language Support, Hooks, OpenTelemetry

Copilot SDK v0.3.0-preview.1: Multi-Language Support, Hooks, OpenTelemetry

Copilot SDK v0.3.0-preview.1 brings public preview status with full multi-language support, OpenTelemetry observability, hooks system for permissions and auditing, and session filesystem abstraction. Breaking changes require permission handlers and update Python to 3.11+.

Infrastructure Noise in AI Coding Evals: The 6-Point Leaderboard Gap

Infrastructure Noise in AI Coding Evals: The 6-Point Leaderboard Gap

Anthropic found that infrastructure configuration alone creates a 6-point spread on Terminal-Bench scores — larger than most leaderboard gaps. Resource limits below 3x cause spurious kills; above 3x they help agents solve different problems entirely.