Summary
Riley Brown’s weekly AI agent update covers several platform moves from the major labs in a single episode. The headline feature is a new OpenAI Codex capability that lets the agent read and send iMessages on macOS, demonstrated live with real message threads. The workflow includes a human-approval step before any message is dispatched, and Brown shows it analyzing a 3,251-message group thread and generating communication coaching — a practical illustration of long-context agent memory applied to personal productivity.
The episode also compares three emerging “super app” platforms — Grokbot (xAI), ChatGPT Work (OpenAI), and Claude Co-work (Anthropic) — across dimensions including ease of use, in-app browser quality, model capability, and iOS experience. Brown gives Grokbot points for simplicity and ChatGPT’s Codex credit for its built-in browser, while arguing Anthropic’s models retain an edge for knowledge work, document creation, and front-end design. Grokbot’s underlying model is rated significantly behind the other two.
Additional items include Anthropic merging its Claude Design and Claude Code products, the launch of Co-work on Anthropic’s iOS app bundled with a structured learning course, and a new Slack integration that embeds Codex and Claude directly into workspace channels. Brown frames this wave of updates as a race to become the default agent layer for both personal and professional workflows, with each platform staking out a different UX philosophy.
📺 Source: Riley Brown · Published August 21, 2026
🏷️ Format: Roundup
