Grok Bot and Muse Glimmer land in one week, marking the shift from chat to always-on agent labor

SpaceXAI opened its persistent-teammate beta at $120–$300 per seat on August 11; Meta open-sourced a 30B agentic model that runs on a laptop the day before.

Two agent launches inside 48 hours have effectively closed the chat era. On August 10, Meta Superintelligence Labs shipped Muse Glimmer, a 30-billion-parameter open-weight model licensed Apache 2.0 and tuned to run locally on a Mac or PC with a single consumer GPU. The next day, SpaceXAI (the division formerly branded xAI) opened the beta of Grok Bot, a persistent-teammate product priced from $120 per seat per month on Cursor Teams Premium to $200 on Cursor Ultra and $300 on SuperGrok Heavy.

The two releases sit at opposite ends of the same thesis: agents that keep working when the laptop is closed.

Grok Bot is the enterprise wager. Each Bot runs in its own cloud computer, calls tools across a catalog of 220 plugins (Google Workspace, Microsoft 365, Slack, Salesforce, Notion, GitHub, Jira), and falls back to a browser when no plugin exists. It pauses for human approval before sending, publishing, buying, or deleting, with model-based review of tool calls. SpaceXAI says its own teams already run Bots for sales outbound, marketing campaigns, office operations, and bug fixes, coordinated by a “chief of staff” Bot managing specialists.

The commercial urgency isn’t subtle. The SpaceX S-1 disclosed the AI unit losing $2.47 billion on $818 million of revenue in Q1 2026. Seats are the road to daylight.

Muse Glimmer is the opposite bet. Mark Zuckerberg framed the release in a 6,500-word essay as a direct challenge to Chinese open-weight labs, and Meta is shipping llama.cpp, MLX, and ExecuTorch integrations in the coming days. The model reports results on DeepSearch QA, MCP-Atlas, τ-Bench, and SWE-Bench, and is trained to retry when a tool call fails rather than halt, the agentic detail that matters most in production.

Between the frontier subscription and the local weights sits the platform layer where Glean, Dust, and LemonLime are competing to orchestrate agents against the tools companies already own. That’s where the pricing pressure will actually land. The chat interface was a demo; persistent labor is the product, and the industry has now priced it.

Grok Bot and Muse Glimmer land in one week, marking the shift from chat to always-on agent labor

Sources