On February 28, Humane AI and its AI Pin ceased to exist. Through our category design lens, that outcome long seemed inevitable. Humane AI violated two critical concepts in category creation and development.
- Don’t get way out ahead of the adjacent possible.
- If you’re the only one in a category, you’ve got a serious problem.
Apparently those rules aren’t obvious to a lot of supersmart people. Salesforce CEO Marc Benioff invested early in Humane AI. So did OpenAI chief Sam Altman. Softbank and Tiger Global pitched in. The company raised more than $230 million. They all bought into the idea that AI pins would disrupt and replace smartphones.
But, for starters, Humane AI failed the adjacent possible test.
Early in Humane’s existence, The New York Times described what Humane AI was supposed to be. The two founders, who came from Apple, “set out to create a lapel pin that clips to clothing with a magnet,” the Times wrote. “The device gives users access to an A.I.-powered virtual assistant that can send messages, search the web or take photos. It is complemented by a laser that projects text onto the palm of a user’s hand for tasks like skipping a song while playing music. It also has a camera, a speaker and cellular service.”
Very futuristic, no?
In 2010, Steven Johnson published the book Where Good Ideas Come From: The Natural History of Innovation. It’s a study of the patterns throughout history that lead to an explosive moment when an innovation – such as the pencil, flush toilet or battery – catches fire and changes the way we work or live.
To describe that moment, Johnson borrowed the concept of the adjacent possible from biology and modified it for his purposes. Johnson divides innovations into two categories: the possible and the not-yet-possible. The possible are things that already exist and work and are understood and adopted by the market. The not-yet-possible are, essentially, lab experiments and dreams – technology that doesn’t yet work well and that the broad market hasn’t adopted.
The adjacent possible is the thin band between these two zones. Innovations change the world when they land there. Such innovations stretch the possible beyond where it’s been before, but not so much that the technology doesn’t work or we can’t understand it.
The tech ecosystem, throughout its history, has pumped out “visionary” companies that started out too far into the not-yet-possible, and then ran out of time and money before their products landed in the adjacent possible and became truly useful and cool. General Magic was a famous case. So was Teledesic, which was supposed to be Starlink long before a Starlink was really possible. Today’s quantum computing companies are likely too far into the not-yet-possible.
We love visionary companies for showing us what a future might look like, but if they’re not aware of the adjacent possible and get practical about developing their visions, they’ll be gone before their visions become real.
From the very first demos of Humane’s Ai Pin, it seemed obvious it was too far into the not-yet-possible – obvious it wouldn’t get traction in any near future. The company struggled to make the technology work. And people in general are quite enamored with their smartphones, and so are years away from understanding why they’d want an AI pin to replace those phones.
As a result, Humane found no real market to sell into, and no healthy ecosystem of suppliers, developers and partners. The company started sputtering, top executives left, and the press ate it alive. In early 2025 it closed its doors and sold some of its assets to Hewlett-Packard for $116 million – about one-third of the amount of money investors had put in. And no, HP doesn’t plan to revive the pin or anything like it. It just wants some of Humane’s software and people.
Then there’s Humane AI’s other sin, which can be related to missing the adjacent possible: creating a market category of one.
Even once Humane had its product on the market, no one else stepped in to develop anything like it. The category of “AI pin” had no other members. Most likely, that was because the concept was too far into the not-yet-possible and other companies didn’t see a viable market they could sell into anytime soon. It also could be because no one else thought an AI pin made any sense, even if it worked well.
As the research of economist Paul Geroski revealed, if a company shows us a new category of product or service that really should exist and that the market will desire, you can bet others will join in. They may make different versions or designs of that product or service, but they’ll be in the same category. Tesla pioneered the electric car category. We knew it was real and enduring when most of the world’s automakers and a bunch of startups piled in and said they, too, would make electric cars.
When a category remains a category of one, it’s a sign that the category just doesn’t matter. If you run a startup and it’s creating a new category, the idea that you’d like to have no competition is lunacy. You should hope the category gets populated with competitors. That makes the category come alive and signals to potential customers – and investors – that the category is real and worth buying into.
Visionary companies help the world move forward. But they effectively do that when they’re actually successful. An important factor in getting there is being aware of the adjacent possible and creating a viable category that others want to join.
–
If you want to discuss your category design challenges with CDA, you can book office hours here.