Why are we seeing so much AI hardware?

This is The Startup Breakdown, the newsletter where we learn, laugh, and love startups. By joining this growing community of hundreds of future startup aficionados (think i spelled that right?), you're getting a beachside view of the ocean that is the startup and VC scene. This ain’t your grandpa’s newsletter, so prepare yourself for an inbox full of 4/20 jokes and Succession references.

Happy Thursday, folks.

A lot of you are new around these parts, and because I didn’t send last week, quick refresher on how this normally works:

• I send three newsletters every week, two of them free news segments like those you received each of the last two days

• On Thursdays, I send a longer, more in-depth premium newsletter where I really dive deep into some significant piece of legislation, trend, or decision that has critical importance to the startup ecosystem

Because I was too busy fighting a virus last week, I wasn’t able to punch your inbox, so to make up for it, this is a premium edition for free.

All access memberships are $4.99/mo (or $49.99 annually) and give you access to all Premium newsletters, past and upcoming. In case you want to read a couple to determine whether you’d like to upgrade, here are a couple of past articles:

I’m narcissistic enough that I’d probably keep blessing the world with my rants even if not for the financial benefits of doing so, but your support helps me to justify the hours I put into making this newsletter everything that it is, and it means more than you can imagine.

I nerded out just as much as the next tech bro when I saw rabbit unveil their r1.

When CEO Jesse Lyu “taught” rabbit how to perform specific complex tasks by “watching” him perform them first, I was hooked. Plus, I was impressed with features like:

• calling an Uber

• playing music

• finding recipes based on what’s in the fridge

…but once the initial drooling wore off, I did start to ask myself the question: why isn’t this an app?

Beyond r1, we’ve seen a recent wave of AI hardware devices, from Tab raising $2 million seed for its always-on pendant, Humane grabbing hundreds of millions and launching its AI Pin a few weeks ago, and Meta working on making robo-shades cool since they were still a social media company.

Why would so many startups elect to devote so much time, money, and risk to seemingly reinvent the wheel to build in AI? Is it really necessary?

Turns out, it is.

As I see it, there are going to be two categories of AI: reactive and proactive.

Reactive products are those like ChatGPT. You put in an input, and the model works to provide the optimal response. Most of the current tools are reactive, whether ChatGPT, Perplexity, Midjourney, etc.

Reactive tools help to explain why we’re seeing a divergence in the levels of utility provided by AI between users. 

• ~95% of people use AI tools either infrequently or for low value-add tasks like answering homework problems or simple questions that a Google search fails to adequately address

• ~5% of people are getting really, really productive using AI tools, learning how to solve homework problems, automate tasks, and/or spend more time on high-level work that requires more reasoning than AI can (currently) provide

Most people don’t know how to ask the right questions. Crazy salaries are being thrown at “Prompt Engineers” who are simply good at getting ChatGPT to answer in the way that they want.

In all honesty, you prolly have to be a bit on the spectrum (🙋) to intuitively understand the nuances of texting a robot lol

The tools which will most benefit the 95%ers are those which proactively provide answers to the user without them having to ask a question. It’s simply not possible to teach everyone how to think when they’re interacting with AI.

This was the entire impetus behind our thinking for Infobot. We send users a daily newsletter telling them exactly what they need to know on the topics that they need to know about.

In the case of my last startup, Jupyter, we were attempting to preemptively handle personal financial management as most people simply don’t know what they want and need, impacting their use of available resources and producing inequality in outcomes.

Hardware elevates this human-AI relationship through nonstop interaction.

Take Tab, for example, which doesn’t even position itself as another personal assistant. Instead, it declares that its nonstop watching and listening will enable it to understand the person wearing it better than anyone else ever could.

The thesis is that there’s a limit to the depth of understanding that one human can have with another imposed by an observation barrier. The most that can be understood about another is what you directly observe (<1% of the person’s existence) and what the other person tells you, which is also such a small fraction (and biased recount) of the person’s life.

For example, take a therapy session:

• Psychologists are paid thousands of dollars to be the world’s best at understanding another person

• They can observe every twitch of the lip, every pulled glance, and every word that comes out of the person’s mouth in 1 hour

• For the other 167 hours, their understanding is restricted by what the patient tells them - a broad overview lacking forgotten details and a bias induced by insecurity over the way that the patient is perceived

Tab wants its pendants to become an extension of the wearer, providing the sort of observation required for the most complete understanding of the person. In fact, they claim their business model is as “Friendship as a Service.”

This level of daily integration isn’t possible through an app. Apple would never allow it, no matter how many permissions you give an app. Frankly, they’d prefer to be able to do this themselves, and they’d never support the ability of third-party services to do so and gain a leg up on the data that they collect about their users.

Hardware is essential, hence why we’re seeing such a wave of developments in the space, but the underlying motivation is also reflective of the potential that exists in the proactive AI space.

Founders and investors are willing to bear all of the technical, financial, and timeline risks of building out these devices because they understand that if it works, this is the next age of technology: it becomes invisible.

Adoption is a concern. Many tech enthusiasts who will have no reservations about giving it a try, but I have concerns over whether the general population will sign up for an all-seeing necklace or pin, particularly given the connotations this carries.

Tab’s wont be the last announcement from companies exploring the potential of the hardware space. How much of this potential is realized, however, is certainly another question.

If you’re looking for an outside opinion on the topic, I actually had an Infobot interview with !Boring (pronounced Not Boring) founder and former consumer hardware founder Andy Allen who had a tweet go semi-viral discussing this (at least as viral as you can when engaging with the topic of consumer hardware lol):

@infobotnews

🌟 Embracing the Future of AI! 🚀✨ 👉 Did you know that hardware-enabled apps have a unique power? 📱💪 They go beyond simple directions, like ... See more

The internet has devolved into nothing but noise, and AI-generated content isn’t doing your digital experience any favors. We filter through that, bringing you all of the news related to tech, business, and regulation that you need to know. Subscribe now and stay informed.

Cheers to another day,

Trey