Breaking out of the God Box

The internet used to be fun.

Facebook, now worth $1.5 trillion, started as a hot or not website at Harvard. TikTok, whose parent company is worth over $500bn, started as a fun way to share dance videos.

So why can’t we have more fun with AI?

People are afraid that with over 1 billion users, OpenAI’s God Box has already won.

But every technology this powerful has created new ways for people to play, live and connect. At Greylock, we see two opportunities: the intelligent personal professional, and networks for a new class of creators. We think both can produce generational companies in the next 18 months.

The intelligent personal professional

Imagine the fleet of professionals reserved for the few, available to everyone.

Concierge doctors, personalized tutors, strategic accountants, personal CFOs, wealth managers, executive assistants, chiefs of staff, nutritionists, personal trainers, lawyers, travel agents, house managers, event planners, therapists, career coaches, matchmakers, etc.

This is inevitable. It’s a big part of why ChatGPT reached over 1 billion users faster than any product in history. And OpenAI has a right to win, because having all your personal AIs in one place makes each one better. It helps if your personal trainer knows about your neck injury.

OpenAI’s consumer opportunity is still underestimated by investors, and I suspect by their own team. The god box could be the greatest business ever created, the front door to all consumer commerce. But they're distracted by AGI, enterprise, and building data centers. And one company rarely owns everything.

So the question for entrepreneurs is which use cases break out of the god box?

There are three ways out:

1. Liability absorption. To provide the best end-to-end experience, you take on the licenses, capital and legal risk. An AI bank or doctor needs more than a strong model to get the job done.
2. Purpose-built product. A UI, workflow, or data that lives outside the training set.
3. Enables a two-sided network. Matching problems where liquidity wins. A career coach connected to an AI recruiter. A dating coach connected to a matchmaker.

Here's one view of the use cases most likely to exist outside the God Box, sized by the labor market each one replaces.

Our best guess is the first intelligent personal professionals to break out at scale are the AI primary care doctor and AI wealth, estate, and tax. OpenAI recognizes this opportunity and are pursuing both, with their acquisition of Hiro in finance and launch of ChatGPT health. But both markets are still up for grabs. They are scenario 1 plays, with licenses and regulatory complexity OpenAI doesn’t want to own to nail the end-to-end experience. They also happen to sit on top of the two largest wage pools.

Networks for a new class of creators

Every platform shift lets amateurs create, and the amateurs often win. That’s why YouTube’s $40bn in ad revenue in 2025 is larger than the ad revenue from the 4 major studios, combined.

The businesses that capture this value are the ones that understand abundance, enable new types of creation, and own the two-sided network of creators and consumers.

So what new types of creation does AI enable?

Today, AI evolution is at the equivalent of mobile evolution’s “camera filter” era, not yet its Instagram era.

In other words, there are amazing creator tools that turn natural language into music, videos, and code. But no two-sided networks at scale.

Our best guess is the first network to break out will be a social game platform. Anyone can now turn natural language into games. Musical.ly/TikTok went from ridiculous to bigger than Hollywood in 10 years with user generated videos; user generated games can take over the industry much faster. Music may spawn a network too, but it's harder to imagine consumption changing from Spotify until AI music becomes bigger than traditional, or the consumption format changes.

An AI game network won't just be big. It will be how the next millions of entrepreneurs earn a living as traditional jobs transition to new ones. And built the right way, it enables human connection. Social media has evolved into a feed of memes from people you’ve never met. Social games are a more natural way to hang out with your friends.

The new social game network will have the same components as existing platforms like Roblox: a game engine where anyone can create, a persistent identity and social graph across games, and an economy to reward creators. The key difference is a game engine that’s truly accessible to anyone with an internet connection and an idea, not the semi-professional studios we see on Roblox. The challenge will be finding the right constraints so amateurs can make fun and viral experiences. And, finding the right way to connect experiences so it feels like a cohesive platform rather than a mix of mini-games. If done correctly, this is one of the biggest opportunities in AI; the metaverse where people play, create, and hang out.

While many take the safe route of building the picks and shovels, a few creative entrepreneurs will change the way people live, play and create.