Elon Musk's SpaceXAI team to live-stream building a startup in 72 hours

From September 15 to 17, three workers at Elon Musk's AI firm will attempt to create a company within 72 hours, relying on AI agents. All choices related to product, engineering, and business will be broadcast live.

The venture currently lacks a name, product, or concept. Matt Palmer, Lauren Tan, and Roshan Sadanani will decide on each of these elements live during the broadcast.

What the 72-Hour Build Involves

The Grok Bot Galaxy event takes place at a location in San Francisco and is streamed globally. Each day's sessions occur between 8.30 a.m. and 6 p.m. Pacific time.

The company name has not been announced. Matt Palmer, Lauren Tan and Roshan Sadanani will start from a blank slate during the Grok Bot Galaxy livestream (Sept 15-17) and decide the idea, plan and name in real time as they build.

— Grok (@grok) September 12, 2026

Grok Bot is an agent product from SpaceXAI, distinct from the Grok chatbot that responds on X (formerly Twitter). The firm says it operates as a group of AI agents that users can label, give tasks to, and deploy across applications and websites.

A domain dispute over Grok Bot involved an anonymous owner who demanded $1 million.

The trio will employ the tool for brainstorming, planning, product meetings, and development. Other team members join for dedicated sessions covering sales, customer service, and marketing.

Why SpaceXAI Is Staging the Test

SpaceXAI came into being when SpaceX acquired xAI in February through an all-stock transaction that placed xAI's worth at $250 billion. Later, in August, it purchased the coding company Cursor for $60 billion.

Throughout the year, Musk has voiced ambitious statements regarding Grok, and competing labs have responded in kind. A three-day unfiltered construction is more challenging to put on than a demonstration video.

Next week, the SpaceXAI team will build a company from the ground and livestream it.

“We'll use Grok Bot for every part of the build – from ideation and product development to real engineering work and deployment.” pic.twitter.com/iz5s0R3ZaE

— Sawyer Merritt (@SawyerMerritt) September 10, 2026

This is still a vendor demonstrating its own product, not an impartial test. The relevant metric is how frequently the three individuals direct, correct, and assist the agents during the process.

Inquiries into who takes responsibility for AI agents when they operate independently continue unresolved.

Musk Concedes Grok Is Not the Default Yet

The construction occurs at an inconvenient time for agent offerings. Competitor Anthropic released a study on September 11 detailing misuse of its Claude models, including cyber operations, surveillance, fraud, and conventional weapons activities. Anthropic stated it eliminated the implicated accounts.

Additionally, accounts indicate that each party in the Gulf conflict uses Claude, a point Elon Musk concedes, admitting that Grok is not the primary choice, at least currently.

“I guess Grok is not yet a preferred choice in this arena. Not sure how to feel about that,” he chimed.

Not appearing in a competitor's abuse report does not indicate safety. Labs only disclose what they detect on their own platforms, and Grok Bot has been publicly available for only about a month.

By Thursday, the broadcast ought to reveal what the term 'company' truly means in this context. A functional product with users would have a different impact than just a name and a landing page.