For most people, the morning starts with reaching for the phone. Karen X. Cheng, an American film director with more than 500 million views, has instead built an AI agent that composes and prints a personal newspaper overnight, before she wakes up. “It prints while I sleep. I used grokbot to pull from my calendar and email,” she described it, according to Komando, which documented her project.
What the agent does overnight
Cheng used Grok bot, an agent inside Grok from xAI, and connected it to her own data. Overnight, it goes through her messages and meetings, pulls out the essentials and assembles them into a finished sheet. According to Komando, the lead story of the day might simply be that someone forgot a 10:30 meeting.
The result is one printed overview of the day instead of dozens of notifications on a screen. The input is one person’s private data. This is what sets the project apart from apps that for years tried to serve everyone the same news.
From personal tool to product
Cheng published her agent as a template on the Grok Bot Marketplace, a marketplace run by xAI. There, anyone can publish their own bot and others can install it with one click, and the offer is divided into nine categories, from engineering through marketing to personal use. Her morning newspaper has moved from a one-off experiment to something that people who would never have built it themselves can run.
This is a pattern that repeats across the creator economy. A creator builds a tool and distributes it as software. Cheng has the profile for it. She is an acclaimed director, listed in Inc’s 30 Under 30 and Adweek’s Creative 100, and the author of the “Girl Learns to Dance in a Year” series with more than 100 million views. In 2022 she provided the prompt for the first magazine cover generated with DALL·E 2 for Cosmopolitan.
The marketplace where she published her bot is not the only one of its kind. OpenAI launched the GPT Store on 10 January 2024 and added a revenue-sharing programme for creators who publish their own GPTs. This creates a distribution layer in which an individual with an idea and one agent becomes the supplier of a finished product.
From aggregation to a personal agent
Large apps tried personalised news before and ran into problems. Artifact, an AI news aggregator from Instagram co-founders Kevin Systrom and Mike Krieger, announced its closure in January 2024, saying that “the market opportunity wasn’t big enough to warrant continued investment”.
Cheng goes in the opposite direction. She builds a tool that each person connects to their own data. It is a shift from an editorial team that selects for the crowd to an agent that assembles for one person.
The European layer and access to data
This is exactly where the project enters European territory. An agent that continuously reads an inbox and calendar processes personal data, which in the EU falls under the GDPR. It raises the question of who processes the data and where it ends up when the input is a private inbox and the processor is a model run by a third party. For a tool used by one person, that is a marginal question. For a template that thousands of people install, it stops being marginal.
For creators in Europe who want to follow the same path and release their own agent as a product, processing personal data is part of the brief from day one. A tool connected to someone else’s inbox is also the processing of someone else’s data.
Karen X. Cheng’s project mainly showed one thing. One person can today write an agent for themselves, and the next day release it as a product for others.
Comments