What the Humane AI Pin Launch Can Teach Product Builders

Earlier this month was the launch of the Humane AI Pin, a wearable pin which you can interact with through voice to speak with an LLM. Reviews were brutal. I'm here not to bury Humane, but to avoid their mistakes.

There were a lot of mistakes (the price, the subscription, the messaging) but I think there was one mistake above all, and one we should all heed when building natural language products.

When launching something new, do one thing really well.

Humane took the opposite approach, claiming to be a phone replacement, even having their own "operating system" they call Cosmos. But in trying to do everything, they ended up setting expectations far too high, and they didn't deliver. What kind of OS can't set a timer? What kind of phone replacement can't reliably make a call?

Of course, a narrow focus is the best advice for building any product, but it's doubly true with natural language products, because a lot of these problems come with the state of AI at the moment.

The most advanced language models are expensive to run, so you have a $700 device with a $24 per month required subscription. They're also resource intensive, so you have a device that overheats, chews through batteries, and takes seconds to give a response. Speech recognition and NLG are both super powerful, but prone to errors, so you have more frequent misunderstandings.

Narrow the focus, though, and you can avoid many of these issues. For example, if this was a device that was only for keeping a user's thoughts and ideas throughout the day, then Humane could have focused on improving the speech recognition and building an amazing model for summarizing that day's, week's, and month's notes.

Is there a market for this? I don't know! But the point is that building a small thing that works really well for a small market is a great place to get started. You can always expand later. Build something for everybody that works for nobody is often the end of the road.