Gemini 4 Argon lands, OpenCode 2.0 ships, and agents learn to drive your Mac

1. Gemini 4 Argon is available to developers

Google released Gemini 4 Argon on October 13, calling it its frontier model for complex workflows like real-world software engineering, code migrations, and cybersecurity defense. You can use it through the paid tier of the Gemini API, and Google AI Ultra subscribers can try it in AI Studio and Google Antigravity. The same day, GitHub made it generally available in Copilot for Pro+, Max, Business, and Enterprise users, across VS Code, the Copilot CLI, and the coding agent.

Our take: If you're on Copilot Pro+, this is the cheapest way to try it. Give it a long refactor you've been putting off.

2. OpenCode 2.0

Anomaly shipped OpenCode 2.0, three months after the public beta. The headline change is a shared background service that connects the terminal, web, and desktop interfaces. It also adds a rebuilt API, an embeddable SDK (@opencode/sdk), and background subagents and shell commands. Heads up: 1.0 plugins don't work on 2.0, and upgrades aren't automatic. Run opencode upgrade --major.

Our take: The SDK is the interesting part. You can now embed a real coding agent inside your own app.

3. Agent-Native (beta) from Builder.io

Builder.io launched Agent-Native in beta on October 14. It's an open-source TypeScript framework for building AI agents with intuitive UIs. You can run it locally or build online with Builder.io.

Our take: Worth a look if your agent is stuck living in a chat box.

4. Braus: agents that use your Mac apps

Braus is a free, open-source macOS app that lets agents work in desktop software while you keep using your Mac. You can create a worker inside Braus, or connect Claude Code or Codex through its local MCP bridge. The window-to-text pipeline runs on your Mac, but remote models can receive window text and context, so pay attention to what you connect.

Our take: Start with a small task you can easily check, like the team suggests.

5. Snyk Evo Agentic AppSec

Announced today, Evo Agentic AppSec is a team of security agents that operate Snyk's existing deterministic engines to find, fix, and block vulnerabilities in agent-written code. Fixes are queued for approval, not auto-merged. Some agents are in experimental preview for select customers. The Remediation Agent is in open preview.

Our take: It's enterprise-focused for now, but "verify agent code with something other than the agent" is the right instinct for everyone.

Try this today

Have a second agent review your first one. Before you merge vibe-coded changes, open a fresh session (or a different model) and paste: "Review this diff for security issues: exposed secrets, missing auth checks, unvalidated input, and open database tables. List concrete fixes only." A clean context catches things the agent that wrote the code tends to rationalize away.