Hark

AI assistants must use computers like humans: only 1 in 1,000 sites have APIs

Adcock's thesis for his AI lab Hark is that a useful AI assistant has to operate a computer the way a person does (browser, mouse, keyboard) instead of relying on APIs or MCP. His reason is that he estimates only about 1 in 1,000 websites expose an API.

“And fundamental to that thesis was you got to figure out how to get AI to use computers general purpose. You would never hire an assistant that couldn't use a computer. So you got to be able to give things out to it that can do everything you can do. financial models, book flights, order DoorDash, whatever you need to do. It needs to be able to do it all autonomously, but only 1 in 1,000 websites have APIs.”

Meta's smart glasses are "one of the worst products I've ever bought"

Adcock pans Meta's glasses: they have no network of their own, depend on the phone app, and take a long time to pair. He argues glasses and pendants are edge accessories rather than the main AI device, says the end state is a brain-computer interface, and doubts glasses will make Hark's top-10 device list.

“In fact, like the Meta glasses are probably one of the worst products I've ever bought. They're just horrible. They're horrible. Like if you, I, I can't even like figure out how to use it. It doesn't have its own network. It piggybacks on the iPhone network. It means your app needs to be open on your phone. The pairing's long. Like, It doesn't work well. Like, I can't think of any reason why I would need this thing strapped to my head for 14 hours a day. Like, it's just like the wrong device.”

Only build around what's 10x better; 2x won't make anyone switch

Asked how Hark designs beyond existing norms, Adcock says they start from first principles (what has changed, namely AI as a new type of computer). They design only around capabilities that are 10x better than the phone or computer, because a 1–2x improvement won't get people to switch.

“That automation can do a few things that are like, when we were designing this, we wanted to design around key principles that could be like 10x better. If it's like 1 or 2 times better than your phone or computer, you're not going to use it. It's got to be like literally 10x better. What are things now that deep learning brings that are like 10x better?”

Steal thisBefore building a replacement for an incumbent product, list what new tech makes 10x better and design only around those capabilities.

Hark lost a senior AI hire to Meta's guaranteed $36M package

Adcock offered a senior AI-infrastructure engineer from xAI $15–20M in Hark Series A stock, pitching that a couple of 10x's could make it worth hundreds of millions or billions. Meta countered with $36M in RSUs over four years, and the candidate took the guaranteed money.

“And he got an offer to go to Meta for $36 million, 4 years of our shoes. And he was just like, it's kind of guaranteed cash. I go there and I have to weigh this, maybe $200 million at Hark or $20 million or maybe $36 for sure at Meta. And he left and went to Meta and they've been doing that. Every candidate we speak to is making some absurd absurd thing.”