AI is moving beyond chat windows and into the real world. SpaceX wants its employees’ collective knowledge to help train Grok, while an experimental AI manager in San Francisco has recommended firing one of its human workers.

Here’s what you need to know.

Elon Musk told SpaceX employees that the company plans to train Grok using the “sum total” of SpaceX information including direct contributions from its workers.

Musk told employees they would effectively become the AI’s “parents,” with their knowledge, ideas, and beliefs helping shape the system. He also encouraged staff to actively use SpaceX’s internal AI tools and improve them.

And Musk added a much bigger promise: anyone working at SpaceX who wants to travel to the Moon or Mars will eventually get the opportunity.

As AI companies search for higher-quality information beyond the public internet, proprietary company knowledge from engineering expertise to internal processes could become incredibly valuable for training and improving AI systems.

SpaceX has years of specialized aerospace knowledge that most AI companies simply can’t access.

For other businesses, the lesson is simpler: your internal knowledge could become one of your biggest AI advantages.

The best candidate for your next role might not live in the same country. Oyster helps you hire globally in 180+ countries. Payroll, compliance, and benefits included.

Get started with Oyster

An AI manager called Luna, running an experimental San Francisco store from Andon Labs, recommended firing a human employee after they were late for 17 of 23 shifts.

Luna had previously created an attendance policy, then apparently failed to consistently remember and enforce it. Researchers eventually prompted the AI to revisit its policies and evaluate the employee. Luna recommended termination, but human staff reviewed the recommendation and ultimately made the dismissal happen.

This is a fascinating preview of AI agents moving from completing digital tasks to participating in real management decisions.

Luna already handles parts of running the experimental store, including choosing products, pricing, opening hours, branding, and hiring. Now employee management has entered the picture too.

But the experiment also exposed a major limitation: an AI capable of creating a workplace rule couldn’t reliably remember and enforce that same rule over time.