Less than a year after introducing its agentic browser ChatGPT Atlas to the world, OpenAI recently shut it down. My guess is the team decided to make a bet on building an AI browser, but the product wasn’t really working, and Google Chrome proved to be a strong incumbent. That resulted in a decision to shut ChatGPT Atlas down as a separate product, bundle it into ChatGPT, and focus on bets with more momentum.

For different reasons, OpenAI also shut down its video model Sora, even after its viral campaign mimicking Studio Ghibli’s signature images and signing a billion-dollar deal with Disney, because it wasn’t financially viable.

These are just a couple of bets that OpenAI have folded on. It has many other irons in the fire. There’s ChatGPT and Codex. 

There’s also the multi-billion dollar acquisition of Jony Ive’s io, to make a hardware product.

There’s the multi-hundred-billion dollar bet on Stargate, which it partnered with a bunch of other tech companies on. Both of them are running into headwinds; the day before I typed this, Apple is suing OpenAI for stealing trade secrets, and they seem to be having trouble executing on Stargate.

A couple of months ago, I read a phrase in an article about Stargate that caught my eye, “From core technology to partnerships, the company places overlapping bets and abandons all but the most expedient. The approach owes much to chief executive Sam Altman’s ‘venture mindset’, said one person close to the company.” 

I used the word “Bets” earlier, and I like the phrase, “Venture mindset.” It reminds me of all the products that companies like Amazon and Google discontinued, as well as the idea of small bets (though OpenAI’s are not so small). These products are effectively livestock; OpenAI starts or acquires many of them, and only sustains the ones with the most momentum.

OpenAI is often compared with Anthropic, which I see as much more focused. I don’t know Anthropic for much more than Claude, Claude Code, and Claude CoWork, as well as more recently Mythos and Fable. It’s pulling ahead on enterprise sales. It’s renting space from SpaceX instead of building its own. I’m sure there are setbacks, of course, but publicity for Anthropic seems more tightly buttoned up.

The paradigms driving these two director competitors feel different. It’ll be interesting to see how things turn out.