GitHub has shipped a substantial capability expansion for its Copilot cloud agent — the AI coding assistant formerly known as the "Copilot coding agent" — with three new features that push it meaningfully up the software development lifecycle. The first is a branch-first workflow: Copilot can now work on a

OpenAI shipped Codex CLI v0.119.0 on March 31, and the release is packed with improvements across security, usability, and developer workflow integration. The headline security change is proxy-only networking enforcement for Windows sandbox runs — instead of relying on environment variables alone, egress is now restricted at the OS

Microsoft has issued a formal public denial that GitHub ever intended to run advertisements in pull requests — calling last week's Copilot "product tips" incident a "programming logic issue" rather than a deliberate ad product. When Melbourne developer Zach Manson's team asked Copilot

GitHub has officially announced that Claude Sonnet 4 will be deprecated across all Copilot experiences on May 1, 2026. The cutover covers the full breadth of Copilot integrations: Copilot Chat, inline edits, ask mode, agent mode, and code completions. The recommended migration path is Claude Sonnet 4.6. For individual

TechCrunch's deep-dive into OpenAI's $122 billion funding close surfaces the financial architecture behind the headline number. SoftBank co-led the round alongside Andreessen Horowitz, D.E. Shaw Ventures, MGX, TPG, and T. Rowe Price, while $3 billion came directly from retail investors via bank channels — an unusual

OpenAI has closed its largest funding round ever — $122 billion at an $852 billion valuation — with Amazon, Nvidia, SoftBank, Microsoft, and a wave of retail investors among the backers. Alongside the record capital raise, the company disclosed that ChatGPT now serves 900 million weekly users and that its ads pilot

In a move that blurs the lines between competing AI coding ecosystems, OpenAI has released codex-plugin-cc — a plugin that lets Claude Code users invoke Codex directly from within Anthropic's coding agent. Built by OpenAI engineer Dominik and published on GitHub, the plugin supports three workflows: a standard Codex

Microsoft is expanding Copilot Cowork through its Frontier early-access program, and the headline addition is Claude: Anthropic's model is now integrated for long-running, multi-step agentic tasks alongside GPT-4o. The update also ships an improved Researcher agent and a new Critique feature with an explicit multi-model handoff baked in

Developer Zach Manson discovered something unsettling: GitHub Copilot had quietly started editing pull request descriptions to insert promotional text — advertising tools like Raycast, Slack, Teams, and various IDEs — without any user consent. After a teammate used Copilot to fix a simple typo in a PR, the tool also dropped in

Bluesky's former CEO Jay Graber and CTO Paul Frazee took the stage at ATmosphere 2026 this week to unveil Attie — an AI assistant built on Anthropic's Claude that lets anyone create custom social feeds on Bluesky using plain English. No code required. Point Attie at what

A quiet but consequential split is happening inside organizations everywhere. On one side: engineering teams deploying full agent fleets in production, writing code almost entirely with LLMs. On the other: teams still in evaluation mode, running sparse experiments with no agents in sight. The gap is widening fast. McKinsey'

OpenAI has rolled out a plugin system for Codex, its agentic coding assistant, bringing one-click installable packages that bundle workflow prompts, MCP server integrations, and third-party app connections into a unified marketplace. Out of the gate, the plugin ecosystem includes connections to GitHub, Gmail, Box, Cloudflare, and Vercel — dramatically expanding