No product yet. Seventy employees. And yet Nvidia, Qualcomm, AMD, and Salesforce all co-invested, handing over $700M in a Series A at a $6B post-money valuation. So what exactly is this company building?
What are they building?
Hark is an AI lab founded by Brett Adcock in late 2025 — the same person who previously started Figure AI (humanoid robots) and Archer (electric aircraft). This time, he's going after the AI interface.
There's still no product. But Adcock put up $100M of his own money at the start, and recruited Abidur Chowdhury — the Apple industrial designer who led the iPhone Air's design — to run design. The team also includes former Meta AI researchers and designers from Apple and Tesla.
"I haven't seen anything that feels like something that will really help the normal person."
— Abidur Chowdhury, Director of Design at Hark (formerly Apple)
Hark's core concept is the "intelligence base layer." Every AI tool today is an app — ChatGPT is an app, Claude is an app. Hark thinks that's fundamentally wrong. As Chowdhury puts it: "There's so much that we could be doing if intelligence was at the base layer." AI shouldn't be one more app — it should be the infrastructure everything else runs on.
The tasks Hark wants to automate aren't glamorous: form-filling, travel booking, home renovation planning. The idea isn't that AI handles your to-do list — it's that AI understands your context and takes care of things before you even ask.
Why is this different?
Here's the interesting thing about the 2026 AI landscape: the market leaders are all pointing at developers. Anthropic is going after developers with Claude Code, and OpenAI is focused on enterprise software as it prepares for its IPO. Chowdhury called this out directly: "Few companies are focused solely on building interfaces and native hardware the way Hark is."
| Company | AI Strategy | Primary Target |
|---|---|---|
| Hark | Base layer AI + dedicated hardware | Everyone (normal people) |
| Anthropic | Coding tools (Claude Code) | Developers |
| OpenAI | Coding + enterprise software | Developers & enterprises |
| Meta | AI integrated into existing glasses | Consumers (platform constraints) |
Here are Hark's three key differentiators.
- Built from scratch, end to end
Speech, text, and vision models all designed from the ground up — no API stitching. Just as Apple designs hardware and software together, Hark is building models and devices simultaneously. - Explicitly rejecting wearables
Chowdhury said it plainly: "I don't think it's appropriate to put a layer between humanity and the interfaces we use in the world." No glasses, no pins — AI that blends naturally into the screens and surfaces already in your life. - Personalized for each individual
Not "the simplest thing for everyone" but "the right thing for each person" — that's the design philosophy. Persistent memory systems build a genuinely customized experience over time.
There's an obvious unsolved problem, though. For AI to understand your daily life, it has to collect a lot — and protecting the privacy of bystanders who haven't consented is something Hark itself admits has no clean answer yet.
Nothing exists yet
Hark raised $700M before shipping a single product. First AI models are expected summer 2026; hardware devices come after. Whether the $6B bet pays off is something only time will answer.
How to start: the practical guide
You can't try Hark yet. But getting ahead of this shift is worth the prep work.
- Apply for beta
Head to hark.com and sign up for platform access. Early access is available before the models drop — summer 2026 is the expected window. - Understand "app AI" vs. "infrastructure AI"
Getting fluent in the base layer concept helps you track where this space is heading and spot the product opportunities it creates. - Actually use competing hardware AI
Try Meta AI glasses or Android spectacles firsthand. You'll develop a real intuition for the problems Hark is trying to solve differently. - Map your own "boring repetitive work"
Travel booking, form-filling, repeat emails — that's exactly Hark's target. Knowing your own list means you're ready to apply these tools the moment they arrive. - Get clear on your privacy stance
How much context are you willing to share with AI? This conversation is about to get serious. If you're building products, start thinking about how you'd explain this to users.
