Descriptions:
Matthew Berman provides a hands-on walkthrough of Grockbot, a new agentic platform jointly developed by Cursor and SpaceX AI that aims to bring multi-agent workflows to a significantly broader audience. Unlike Cursor or Codex — which surface tool calls, directory trees, and code in real time — Grockbot deliberately hides implementation details behind a chat interface modeled closely on Telegram or WhatsApp, with the stated goal of making agent-driven work accessible to non-technical users.
Each bot in Grockbot runs inside its own persistent virtual computer spun up in the cloud, with shared authentication across agents so users only sign in once per service. Berman demonstrates the inter-agent communication system, where a chief-of-staff bot delegates tasks to specialized agents (email, calendar, research) and the full conversation history between agents persists and is visible to the user. He also shows the local computer control capabilities, allowing bots to query and interact with the user’s physical machine in addition to the cloud environment — a hybrid model he describes as notably different from the default-local approach of Claude Code and Codex.
Berman flags one design tension: Grockbot’s thread-per-agent organization feels less intuitive than a single threaded conversation for users accustomed to topic-based chat. Overall the review positions Grockbot as a meaningful step toward mainstream agentic computing, though its suitability for technical power users versus the non-technical audience it targets remains an open question.
📺 Source: Matthew Berman · Published August 12, 2026
🏷️ Format: Review
