Blog
xAI explains how it designed Grok Bot for persistent, multi-agent work
On September 3, 2026, SpaceXAI published a design essay on Grok Bot explaining why the product is organized around a Bot roster rather than disposable chat history. The post covers five interface primitives, presence and avatars, each Bot's own computer with takeover, Routines, and practical roster limits.
Designing for agents that outlive a session
On September 3, 2026, SpaceXAI published a design essay on how it built Grok Bot for agents that persist beyond a single chat session. The piece explains interface choices for bots that keep identity, memory, and responsibility over timeincluding work that can start without a user prompt.
Most AI products organize around a chat session the user operates. Grok Bot was designed for an agent that can carry responsibility on its own, which meant reconsidering the sidebar, progress signals, and when work becomes visible.
Five primitives
AI products have accumulated many overlapping concepts. SpaceXAI argues that exposing each one asks users to understand more than they need.
The team kept five primitives:
- Bots — persistent agents with their own identity, memory, runtime, and tools
- Chats — the conversational interface for working with a Bot
- Prompts — context or instructions; usable once, saved as Skills, or triggered as Routines
- Tools — access through software, APIs, connectors, the shell, or computer use
- Artifacts — durable outputs such as documents, designs, code, and data
Everything else stays beneath the interface until the user has a reason to care. The next question was which object should organize the product.